Every reader who arrives at John Roemer’s work on exploitation carries a picture formed by the classical account, and the picture is worth naming before anything is disturbed. In that picture, exploitation happens inside production. A worker sells the capacity to labour for a wage, works longer than the time required to reproduce the value of that wage, and hands the remainder to an employer who owns the product because he owns the workplace and its instruments. The wrong, on this telling, is something that occurs during the working day, between particular people, in a relation one of them directs and the other endures. Roemer’s reformulation reaches the same word from the opposite direction. It never enters the factory, the office, or the field. It asks a question about the distribution of productive assets with which people arrive at the economy’s door, and it derives every verdict about who exploits and who is exploited from that distribution alone.

A reading desk with the two accounts compared table and the withdrawal test worked in longhand, Insight Crunch

The distance between those two starting points is the subject of this article, and it is larger than it first appears. A definition that begins from ownership and a definition that begins from the labour process do not merely emphasise different things. They disagree about where the phenomenon lives, about what evidence could establish it, about which features of a workplace matter and which are decoration, and about what would have to change for the phenomenon to disappear. Roemer, an economist formed inside mathematical general equilibrium theory and writing in sustained engagement with Marx, pressed the ownership starting point with unusual rigour. He stated a test that any reader can apply without equations, proved that class positions and exploitation positions coincide under conditions he specified exactly, and then proved a further theorem whose consequence still startles readers encountering it for the first time: the formal argument that is supposed to single out labour as the exploited factor goes through unchanged when another commodity is put in labour’s place.

How does Roemer define exploitation?

by a withdrawal test. A group is exploited if it would be better off leaving the economy with its per capita share of society’s productive assets. Exploitation therefore follows from unequal asset ownership rather than from anything occurring during the working day.

Hold the shape of that answer in view, because everything else in the article is an unpacking of it. The test is counterfactual: it does not describe anything a group has done, nor anything done to it, but compares its actual position with a hypothetical position it would occupy after a withdrawal that never takes place. The comparison turns on assets, meaning the produced means of production and other alienable resources an economy contains, divided by heads. A group that would gain from leaving with an equal per-person slice of those assets is exploited in the actual arrangement. A group that would lose from such a departure is exploiting. A group indifferent between staying and leaving stands in neither relation. Nothing in the test mentions hours, effort, supervision, wages, or the interior of any enterprise, and that silence is deliberate rather than accidental.

Exploitation relocated: from the working day to the asset distribution

The relocation named in this article’s governing claim is a movement of the concept’s foundation. In the classical architecture, the foundation is a process: the extraction of surplus labour inside production, analysed in the account of the classical concept of exploitation, which this article treats as given and does not re-derive. In Roemer’s architecture, the foundation is a state of affairs: an unequal division of the economy’s productive wealth among the people who must live by deploying their labour and whatever else they own. Process gives verdicts about episodes and relations. State of affairs gives verdicts about distributions and the positions they generate. Roemer’s wager was that the second kind of verdict could reproduce everything the first kind was wanted for, class structure included, while resting on premises a sceptical economist could inspect one by one.

Where does the classical account locate the wrong of exploitation?

Inside production itself, in the gap between the labour a worker performs and the labour embodied in the wage goods the worker receives. The employer appropriates the difference because ownership of the workplace lets him direct the process and claim its product, so the wrong is relational and occurs between identifiable parties during the working day.

Roemer’s dissent from that location was not a denial that such gaps exist. His early work engaged the classical machinery on its own ground, and the result known as the fundamental Marxian theorem, proved in related forms by Michio Morishima and others, established that positive profit in a capitalist economy corresponds to positive surplus labour under stated conditions. The dissent began when Roemer asked what that correspondence actually certifies. A correspondence between profit and surplus labour shows that two magnitudes move together. Whether the labour magnitude is the morally load-bearing one, the one whose presence makes the arrangement objectionable, is a further question, and Roemer came to argue that the further question answers itself against labour once the same style of argument is run for other commodities. That argument belongs to a later section, and it is the centre of the whole construction.

What changes when the starting point is ownership rather than the working day?

The evidence changes first. Instead of reconstructing hours and values inside production, the analyst inspects who owns what at the outset and computes how positions would shift under an equal division. The moral charge also changes, attaching to the distribution people start from rather than to conduct inside firms.

Behind that change of evidence sits a change of explanatory ambition. The classical account explains exploitation and profit together, through one mechanism, and treats the explanation as the indictment. Roemer’s account separates the two tasks. Explaining profit remains a job for price theory, and economists of many persuasions can share its results. Indicting an arrangement becomes a job for a theory of justice in holdings: if the starting distribution is indefensible, then the positions it generates inherit the indefensibility, whatever courtesies are observed inside the workplace. Readers who feel that something has been lost in this separation are feeling exactly what the Marxist critics of the account later named, and their objection receives a full hearing below. Readers who sense the gain, a charge statable without first winning every battle over value theory, are feeling what drew Roemer himself along this path.

The path had a biographical logic worth a paragraph. Roemer came to Marxism as a trained mathematical economist, and his first book on the subject, published in 1982, attempted to rebuild the theory’s core claims inside the modelling idiom he commanded. The rebuild was an act of loyalty conducted by unfamiliar means. Where the classical texts argued through the categories of value, Roemer argued through agents endowed with assets, choosing how to deploy them under constraints, in economies specified tightly enough that conclusions could be proved rather than asserted. The idiom alienated some of the audience the book most wanted to persuade, a cost examined later in this article. The gain, as Roemer saw it, was discipline: every step in the argument had to survive being stated as a premise another economist could accept or refuse on stated grounds.

Why did an economist rebuild a Marxist concept inside general equilibrium theory?

Because proof was the currency of the discipline he worked in, and claims about exploitation circulated there as slogans rather than results. Rebuilding the concept inside models of optimising agents with differing endowments let each step be inspected, refused, or accepted on stated grounds, which he regarded as a service to the tradition rather than a defection.

That service had a polemical edge. Marxian economics in the period faced a standing dismissal from mainstream practitioners, who treated its central claims as unproven and perhaps unprovable. A reconstruction that delivered theorems changed the terms of the dismissal. A theorem can be attacked at its premises, and premise-attack is productive argument, whereas slogan-attack merely repeats allegiances. Roemer’s models also produced results nobody had ordered, most famously the generalised theorem at this article’s centre, and a reconstruction that surprises its own author carries a credibility that apologetics cannot manufacture. The author followed his results where they led, including to the conclusion that the labour theory of value was dispensable for the ethical argument, a conclusion many readers of the classical texts experienced as betrayal and which he presented as the reconstruction’s honest yield.

One discipline governs everything that follows, and stating it early prevents a family of misreadings. The withdrawal test is a definition offered for assessment, not a proposal that any group actually secede from its economy carrying machinery and land on its back. Nobody is invited to picture caravans of equipment crossing a border. The counterfactual is an instrument, in the way that a thought experiment in physics is an instrument: it isolates the contribution of one variable, the distribution of assets, by imagining that variable equalised while the people and their capacities remain as they are. What the instrument measures is whether a group’s disadvantage is attributable to the asset division rather than to its members’ choices, talents, or luck. Verdicts delivered by such an instrument stand or fall with the instrument’s design, and the design choices embedded in it, which assets count, which group is tested, what the withdrawing group is imagined to take, are where the real argument lives.

The withdrawal criterion: a test stated as a counterfactual

Strip the test to its skeleton and it has three moving parts. A group of people is selected. Their situation inside the existing economy is compared with the situation they would occupy if they withdrew from it carrying a specified endowment. The comparison’s direction delivers the verdict. Roemer developed this skeleton in the 1982 book and refined it across the essays gathered in Free to Lose in 1988, and the skeleton’s elegance is easy to miss because each moving part conceals a decision. Which group may be tested: any coalition, from a single household to half the population. What the departing group carries: this is the variable Roemer adjusted to generate definitions for different social forms, and the adjustments are where the theory’s historical reach comes from. What counts as better off: the group’s own valuation of its consumption and leisure under the counterfactual arrangement, assessed by the model’s standard of informed preference.

Why is the withdrawal test stated as a counterfactual?

Because the contribution of asset inequality can only be isolated by imagining it removed. A group’s actual income mixes the effects of its endowment, its effort, its talents, and its luck. Comparing that income with the income the same group would command after an equalising departure separates the endowment effect from everything else the members bring.

Consider what the comparison quietly assumes. The departing coalition is pictured organising production among its own members with the assets it carries, in an economy otherwise resembling the one it left, at prices resembling those it left behind. The group neither gains nor loses trading partners in a way that would swamp the comparison. These are idealisations, and Roemer was explicit that they are idealisations, adopted because the question being asked is itself idealised: is this group’s position explained by the unequal division of productive wealth, or would the group occupy roughly the same position under an equal division? A test that tried to simulate an actual secession, with collapsing supply chains and stranded specialisations, would answer a different and less interesting question, the logistics of departure rather than the justice of the starting point.

The criterion also carries a less obvious assumption about what exploitation is a property of. In the classical picture, exploitation characterises a transaction or a relation: this employment contract, this working day. Under the withdrawal test, exploitation characterises a group’s position relative to a distribution, and the same transaction can change its moral description when the background distribution changes. A wage bargain struck by a worker with substantial assets of her own, who could withdraw comfortably and chooses employment for the sociability of it, does not register as exploitation under the test even if surplus labour is performed inside it, because the worker would not gain from the equalising departure. A bargain struck by a propertyless worker registers even if the employer is courteous, the hours short, and the canteen excellent. Readers who find the first verdict perverse have located, early, the nerve of the Marxist objection developed later in this article.

Which assets count when a group imagines leaving?

Alienable productive assets count for the capitalist form of the test: land, machinery, structures, financial claims on produced means of production, the stock an economy can in principle redistribute by dividing it among heads. Skills and talents are treated separately, as inalienable endowments, and they generate a distinct form of the test rather than entering the capitalist calculation.

The alienable and inalienable distinction does heavy work. Alienable assets can be imagined changing hands without changing who anyone is; a lathe reassigned to a different owner remains the same lathe. Skills cannot be reassigned in that way, since my training cannot be lifted out of me and dealt to another person like a card. Roemer’s taxonomy exploits the distinction systematically. Capitalist exploitation is defined by withdrawal carrying an equal slice per person of the alienable stock. A different social form, which he called socialist exploitation, is defined by a hypothetical withdrawal with inalienable assets likewise dealt out per head, a counterfactual that cannot be physically enacted but can be posed, and under which people of scarce talent would be found exploiting the less talented even in an economy where all produced means of production were held in common. The taxonomy thereby generates a map of successive historical complaints: each social form abolishes one layer of the inequality that defined exploitation under its predecessor and leaves a further layer standing, which the next form’s test then exposes.

Feudal exploitation occupies the taxonomy’s first cell, and its test differs instructively from the capitalist one. Under feudalism, the relevant withdrawal carries the group’s own assets rather than an equal share of the whole stock, because the feudal grievance was not that land was divided unequally in the abstract but that the peasant household, working its own holding with its own tools, owed labour or dues on the lord’s demesne as a condition of its tenure. A group of peasants is feudally exploited if its members would be better off departing with the assets they already possess, freed of the obligations the feudal relation imposes. The test registers the compulsory transfer itself, not an unequal starting division, because the feudal relation’s injustice, as Roemer reconstructed it, lay in the enforced obligation rather than in the distribution of holdings. The contrast between the two tests is the taxonomy in miniature: vary the load the departing group carries, and one formal instrument pronounces on a different social order.

How does the test separate feudal from capitalist exploitation?

By the endowment the departing group carries away. Under the feudal variant a group takes the holdings already in its hands, and release from feudal obligations alone improving its lot is enough for the verdict. Under the capitalist variant the group takes an equal portion per head of all alienable wealth, so unequal ownership itself is what gets measured.

A fourth cell completes the scheme, and it is the one readers of the twentieth century’s history reach for first. Roemer described a form he called status exploitation, arising where position in a bureaucratic hierarchy rather than ownership of assets determines a group’s command over resources. The test adapts again: a group is exploited in this sense when it would gain under a withdrawal in which the privileges attaching to office were levelled. The point of including the cell is not to settle any historical verdict, and this article follows the series discipline of treating the record of states separately from the theory of concepts. The point is architectural. One instrument, adjusted at a single joint, yields a sequence of definitions that between them cover the social forms the Marxist tradition has wanted to criticise, and the sequence displays the tradition’s own narrative of progress and residue: feudal obligation abolished, capitalist asset inequality exposed; asset inequality abolished in imagination, skill inequality exposed; skill inequality posed as the next grievance, bureaucratic privilege waiting behind it.

There is a temptation, meeting this taxonomy, to admire the instrument and forget to ask whether its readings are true. Roemer resisted the temptation in print, and the article returns to his own hesitations when it reaches the counterintuitive verdicts he catalogued. Before those, one further design decision deserves attention, because students of the account stumble over it more than any other. The test is applied to groups, and the verdict for an individual is inherited from the verdict for the groups the individual belongs to. A person belongs to many candidate coalitions at once: her occupation, her region, her asset class, her household. Some of those coalitions would gain from withdrawal and others would not, and the theory must say which coalition’s verdict defines her status. Roemer’s answer, in the capitalist application, runs through the class-exploitation correspondence examined in the next section: the economically operative grouping is the one generated by the asset distribution itself, as agents sort into hiring, working for themselves, and selling their labour. The grouping is thus not chosen by the analyst’s convenience. It falls out of the model.

What would a withdrawing group need to know about itself?

Its members’ capacities and preferences, the assets it would carry, and the terms on which its members would organise production among themselves after departure. The test evaluates the group’s position by its own lights, so a group that values leisure highly may decline a departure that raises its consumption while lengthening its working time.

That last clause deserves emphasis, since it protects the criterion from a crude misreading. Better off is not a synonym for richer in goods. A coalition of people who prize free time, and who inside the existing economy have arranged short hours at modest consumption, might find that withdrawal with an equal asset share would raise their consumption only if they worked longer than they wish to work. Judged by their own preferences, the departure does not improve their position, and the test does not pronounce them exploited merely because their consumption is low. Conversely, a group whose members work punishing hours for modest consumption, and who would gain on both dimensions after an equalising departure, receives the exploited verdict with the full force of the comparison behind it. The criterion is welfarist in this limited sense: it trusts the coalition’s own sense of what its life is worth, rather than imposing an external schedule of needs. Critics who worry that preferences adapt to deprivation, that the long-oppressed may cease to want what they have never had, raise a genuine difficulty for any preference-based test, and Roemer’s framework has to absorb that difficulty rather than dissolve it. The difficulty returns in the objections section, wearing a different coat.

It is also worth pausing on what the test does not require the analyst to know. No reconstruction of embodied labour times is needed. No prices need to be transformed into values, no aggregate equalities defended, no rate of surplus value computed for any industry. The evidentiary basis of an exploitation claim shrinks to an ownership distribution and a counterfactual computation. For economists trained to distrust the classical machinery, this shrinkage was the account’s chief attraction, and it explains the uptake the formalisation found in work on inequality and property rights by scholars with no allegiance to the tradition that produced it. Whether the shrinkage discards the phenomenon along with the machinery is the question the rest of the article adjudicates.

Counterintuitive verdicts and the boundary of the concept

A definition earns trust partly by the cases it handles awkwardly, provided the awkwardness is admitted and priced. Roemer catalogued the awkward cases of his own criterion with a candour that disarmed some critics and armed others, and any honest exposition must walk through the same cases. Three recur in the literature and in seminar rooms. The first concerns choice. Two people begin with identical endowments in an economy where assets are equally held. One works diligently, saves, and accumulates; the other prefers leisure and consumes as she goes. Years pass, and the second person, now asset-poor, must sell her labour to the first, who hires her. Is the second exploited by the first? The withdrawal test says no, provided the asset distribution at the moment of testing, unequal though it has become, arose from choices made under an initially equal division, because the test evaluates the present distribution against the counterfactual of equal division at the present moment, and the diligent person’s advantage traces to a history the criterion does not indict. Some readers find this exactly right: exploitation should not name the wages of idleness. Others find that the criterion has quietly changed the subject from relations between people to snapshots of distributions, since the employment relation between the two people looks, from inside, identical to any other.

The second case concerns the composition of the tested coalition. Verdicts under the criterion are relative to the group selected, and a large coalition’s verdict can reverse the verdicts of the smaller groups inside it. A region of skilled, asset-moderate workers might gain from withdrawal, while the same workers, pooled with a much poorer region into a single national coalition, produce a pooled verdict that conceals the internal transfer. Roemer’s treatment insists that exploitation claims are therefore incomplete until the coalition is specified, and that many public arguments about exploitation trade on unspecified coalitions, sliding between groupings as the rhetoric requires. The insistence is a genuine analytical contribution, whatever one thinks of the criterion itself: it forces any claimant to say who, exactly, is held to be exploited, and compared against which departure.

The third case is the one Roemer himself flagged as the criterion’s most uncomfortable product. Under the socialist form of the test, in which inalienable assets are also imagined equalised, people of scarce and highly rewarded talent emerge as exploited by the untalented majority whenever the talented would gain from a hypothetical world in which endowments of skill were dealt out evenly. The verdict follows from the machinery, and Roemer accepted it as the price of consistency, arguing that a society committed to equality must indeed decide how far the claims of the talented extend. Critics from within the Marxist tradition found the verdict a reductio: a concept of exploitation that can be turned, without alteration, against the least advantaged by the most advantaged has lost contact with the phenomenon the tradition named. Roemer’s reply, in essence, was that the contact was never with a phenomenon but with a family of complaints about unequal endowments, and that the family resemblance among those complaints is what his taxonomy maps. The exchange previews the domination objection, which presses the same discomfort with greater theoretical force.

Why does the choice of coalition matter to the verdict?

Because the test compares a specified group against its own departure, and different groupings of the same people yield different comparisons. A coalition pooled widely can show no gain from withdrawal while a narrower coalition inside it would gain substantially, so an exploitation claim that never names its coalition has not yet made a checkable assertion.

Underneath the three cases sits a distinction the criterion enforces with unusual clarity, and which students should carry away even if they reject everything else in the account. Being badly off and being exploited are different conditions, and neither entails the other. A person can be badly off without being exploited: consider someone whose poverty flows from a disability that no redistribution of alienable assets would relieve, in an economy whose asset division is equal; the withdrawal test finds no gain from departure, because departure changes the asset division and the asset division was not the source of the disadvantage. A person can also be exploited without being badly off in any absolute sense: a well-paid employee of a profitable firm, asset-poor by choice of consumption, may sit inside a coalition that an equalising departure would benefit, and the test returns its verdict without consulting her comfort. The classical account enforces a parallel distinction in its own vocabulary, since surplus labour can be extracted from workers who are, by historical standards, prospering. Both accounts thus refuse the journalistic habit of using exploitation as an intensifier for poverty. They refuse it for different reasons, and the difference in reasons is the relocation once more: one refusal rests on whether a process of extraction is present, the other on whether a distributional counterfactual holds.

How does the criterion distinguish exploitation from mere hardship?

The source of the disadvantage. Exploitation, on the withdrawal account, names disadvantage that an equal division of productive assets would remove. Badness of condition can arise from sources no asset division touches, such as illness or misfortune, and comfort can coexist with exploitation when a well-paid but asset-poor group would still gain from departure.

There remains the question of how a departing group is to treat trade with the economy it left. In Roemer’s constructions, the departed group is typically imagined as autarkic, reproducing among its members the activities it previously conducted with outsiders, and the comparison is between its autarkic welfare and its integrated welfare. Autarky is a modelling convenience with teeth. Real groups are specialised, and specialisation means that departure destroys value that integration created, so a group can fail the withdrawal test, can be pronounced unexploited, partly because departure would strand its specialised skills rather than because its position is just. Roemer recognised the point and treated the test as measuring exploitation net of the genuine gains of cooperation, which is defensible: no theory of justice should count the benefits of cooperation itself as a wrong done to the cooperators. Yet the defence has a cost. The more value integration creates, the fewer groups will gain from the imagined autarkic departure, and the quieter the criterion’s verdicts become, until in a sufficiently productive and interdependent economy almost nobody is exploited by its lights however the assets are divided. Whether that quietness is the criterion’s wisdom or its failure is a judgement the reader should postpone until the objections have been heard in full.

The two accounts compared

Set side by side, the classical account and the withdrawal account answer the same five questions differently, and the table below fixes the contrast in a form that can be consulted while the rest of the article argues over it. Read the rows in order. The first row states where each account locates the phenomenon. The second states what each requires to be true before it will pronounce. The third states the range of what each can explain. The fourth draws out what each implies for the ethical assessment of an economy. The fifth, the row on which the article’s verdict eventually turns, states what each cannot capture, the residue that escapes its net. A table of this kind is a map, not an argument, and maps flatten; the prose sections before and after it restore the gradients. But the flattening has a use. It prevents the commonest error in discussions of this material, which is to compare the classical account’s strengths against the withdrawal account’s weaknesses, or the reverse, instead of comparing strength against strength and weakness against weakness in matched rows.

Question Classical account Withdrawal account
Where is exploitation located? Inside production, in the extraction of surplus labour during the working day In the distribution of productive assets from which agents start
What does it require? A labour process whose performed hours outrun the labour content of the wage goods A coalition that would gain by withdrawing with a per capita share of alienable assets
What does it explain? Profit and exploitation together, through a single mechanism of appropriation Class position and exploitation together, through the sorting of agents by endowment
What does it imply ethically? That the wage relation itself, however courteous, embodies a wrongful transfer That the initial distribution of assets, however courteously administered, is indefensible
What can it not capture? Why the transfer matters if the starting distribution were just What happens between people inside production, including direction and obedience

Linger on the fourth row, since it governs the political consequences drawn from each account. If the wrong is the transfer inside production, then reform that leaves the wage relation intact leaves the wrong intact, and the classical tradition’s suspicion of purely distributive remedies follows from its definition rather than from temperament. If the wrong is the starting distribution, then a remedy that equalises starting positions addresses the wrong directly, even if wage relations persist afterwards among equally endowed citizens, and Roemer’s later work on market socialism follows from his definition with the same logic. Definitions in this territory are never innocent. Each account builds its remedy into its diagnosis, and part of assessing the relocation is noticing that the relocation of the concept is simultaneously a relocation of the cure, from the organisation of production to the constitution of ownership. Readers who suspect that the argument is won or lost at the moment the definition is chosen have understood something real about theoretical politics, and the suspicion applies symmetrically to both columns of the table.

The third row conceals a subtlety that the correspondence principle, examined next, brings into the open. The classical account explains profit and exploitation together because, in its architecture, profit just is realised surplus value, so the two magnitudes share a source. The withdrawal account explains class and exploitation together, since both classifications are read off the same endowment distribution, so the two classifications share a basis. Each account thus purchases a unification, and each unification is purchased at the price of the other’s territory. The classical unification says nothing, by itself, about why the asset distribution is what it is; that requires a further theory, of primitive accumulation and of accumulation’s ongoing dynamics. The withdrawal unification is equally silent about what occurs inside the firms its agents form; that needs a further theory, of authority and effort extraction, which Roemer’s framework does not supply. A complete social theory would need both territories mapped. The dispute between the accounts is a dispute about which map is the master map, the one whose coordinates the other must borrow.

The class-exploitation correspondence: class read off from endowments

The withdrawal criterion defines exploitation. It does not, by itself, say who the exploiters and exploited are in any actual economy, because applying it requires testing coalitions one by one. Roemer’s second result closes that gap, and it is the result that converted many readers who found the criterion merely ingenious. Build a model economy, he proposed, populated by agents who differ only in their endowments of productive assets. Let each agent choose, freely and with full information, how to deploy her labour and assets: she may work her own assets herself, she may hire the labour of others to work her assets, she may sell her labour to work the assets of another, or she may combine these in proportions the model permits. Let prices form as they form in competitive equilibrium. Then observe where the agents land. The class-exploitation correspondence principle states that the landing places and the exploitation verdicts coincide: agents who end up hiring labour are exploiters under the withdrawal criterion, agents who end up selling their labour are exploited under it, and agents who work only their own assets occupy neither position. Class, in the sense of a person’s relation to the labour of others inside production, corresponds exactly to exploitation status, in the sense of position relative to the asset distribution.

Dwell on what the principle does not assume, because the omissions are the result. The model contains no coercion. No agent is forced into the labour market by threat, none is legally bound to an employer, none is deceived about prices or about her alternatives. The model contains no account of what happens inside any firm: no supervision, no effort bargaining, no authority relation, no working day at all, only contracts chosen and executed. The model does not even assume that agents dislike their positions; the seller of labour may prefer her bargain to every alternative open to her, and may be right to prefer it. From endowments and choices alone, with production treated as a sealed box, the correspondence emerges. Roemer’s conclusion from it was characteristically double-edged. The Marxist tradition had always insisted that class position and exploitation travel together. The correspondence vindicates that insistence as a theorem. But the theorem’s proof routes entirely through the distribution of assets, which suggests that the travelling companion doing the work is the distribution, and that the labour relation is the distribution’s shadow rather than its rival.

How do agents with different endowments sort themselves in the model?

By comparing what each course of action pays. An agent whose assets are ample relative to her labour earns most by hiring others to work them. An agent with negligible assets earns most by selling her labour. An agent in between works her own assets alone. Optimisation under differing wealth generates the class structure.

The mechanics deserve a slower pass, because the principle’s persuasiveness depends on the mechanics being ordinary rather than rigged. An agent deciding how to spend her working time faces prices for labour and for the services of assets, and she holds a portfolio: her labour power, which she cannot sell outright but can hire out by the period, and her stock of alienable assets, which she can deploy herself or lease to others directly or through the mediation of a firm. If her stock is large, her own labour is the scarce input in her personal production plans, and she profits by combining her abundant assets with other people’s labour, paying wages below the value the combined operation yields her. If her stock is small, her assets cannot occupy her labour profitably for a full working period, and she profits by selling the surplus of her time to someone whose stock can occupy it. Nothing in this reasoning mentions power. The sorting is arithmetic wearing institutional clothes, and that is precisely Roemer’s point: the class structure of a capitalist economy can be generated by arithmetic of this kind, from a distribution of assets and a set of prices, without any premise about domination entering the derivation.

What does the correspondence principle claim about hiring and selling labour?

That the two classifications coincide under the model’s conditions. Whoever finds it optimal to hire labour is, by the withdrawal criterion, an exploiter, and whoever finds it optimal to sell labour is exploited, so class position determined in the labour market and exploitation status determined by the asset test are the same division drawn twice.

Two qualifications keep the principle honest, and Roemer stated both. The correspondence is proved for a model with specific features, competitive markets and reproducible equilibria among them, and its extension to economies with uncertainty, imperfect credit, and inherited advantage is a matter of argument rather than proof. The principle also runs through a second market in a way that deepens its meaning. Roemer showed that an economy in which agents lend and borrow funds, rather than hiring and selling labour, reproduces the same structure: lenders occupy the exploiting position and borrowers the exploited one, and under stated conditions the two economies are isomorphic, generating identical distributions of welfare and identical exploitation verdicts. The isomorphism matters because it dissolves a possible rejoinder. A defender of the classical account might concede the labour market result while insisting that the labour relation remains the real site of the phenomenon, the credit version being a financial reflection. The isomorphism blocks the insistence: if the entire structure, verdicts included, survives translation into a world where nobody hires anybody and all transactions are loans, then the structure was never about the labour relation’s specific content. It was about differential endowment, expressed through whichever market the economy happens to use.

Why does the credit market version of the model matter?

Because it shows the class structure does not depend on the labour relation’s particular form. An economy of lenders and borrowers, in which no one hires anyone, reproduces the same positions and the same exploitation verdicts, so the generating mechanism is the unequal endowment itself rather than anything distinctive about employment.

The correspondence also reframes a concept the classical tradition handled through narrative. Primitive accumulation, in the classical telling, is the historical process that created a propertyless class compelled to sell its labour, and the telling is a story of enclosures, clearances, and law. Roemer does not dispute the story as history; the series treats the classical account of that process in its own article and this one links rather than retells it. What the correspondence adds is a claim about the story’s logical role. However the unequal distribution arose, violently or peaceably, by theft or by thrift, its present existence suffices to generate class and exploitation positions in the model, and its violent origin is not a premise the exploitation verdict requires. For some readers this is the account’s liberating insight, separating the indictment of a distribution from contested historiography. For others it is the account’s original sin, laundering a history of expropriation into a static fact about holdings. Both readings agree on the architecture. They disagree about whether the architecture is a clarification or an evasion, and that disagreement is a version of the article’s master question, whether the relocation preserves the concept or abandons it.

Class analysis in the hands of Erik Olin Wright took the correspondence seriously as a challenge and a resource. Wright’s own framework, treated in the companion article on Erik Olin Wright’s class analysis, built class locations out of relations to productive assets and then complicated the map with contradictory locations, organisational assets, and credentials. Wright engaged Roemer’s models directly, accepting the demonstration that asset distributions generate class structures while resisting the reduction of class to a single asset dimension. The engagement is instructive for students because it shows the correspondence functioning as good theorems function in a living research programme: not as a conversation-stopper but as a constraint that subsequent theory must either accommodate or explicitly refuse. Wright accommodated it by pluralising the assets. The pluralisation, in turn, raises a question Roemer’s parsimony must answer, namely whether the single-dimension model’s elegance was purchased by excluding the very assets, organisational position and certified skill, through which much actual advantage flows. Roemer’s taxonomy has cells for skill and status, as the earlier section showed, so the pluralisation can be read as an extension of his scheme rather than a refutation of it. Readings multiply. The theorem stands underneath them, unmoved.

What does optimisation have to do with class position?

Everything, in the model. Class position is not assigned by birth, law, or anyone’s command, but emerges as each agent’s best response to her endowment at given prices. The wealthy hire because hiring pays them, the asset-poor sell because selling pays them, and the resulting structure is an equilibrium of choices rather than an imposition.

A reader formed by the classical tradition will want to protest at the word choices, and the protest has a distinguished pedigree that the objections section will give its full weight. For the moment, note only that Roemer anticipated the protest inside the construction itself. The agent who sells her labour chooses to sell it, given her endowment. Change the endowment and the choice changes. The theory therefore locates the unfreedom, if that is the word wanted, not in the transaction but in the constraint set the agent faces when transacting, and the constraint set is the asset distribution wearing another name. Whether unfreedom can be relocated into a constraint set without remainder, whether a person hemmed in by her options is unfree in the same sense as a person directed by a supervisor, is one of the deepest questions the account raises, and it will not be answered by definitional courtesy. It is taken up directly when the objections are heard.

What the correspondence establishes, and what it leaves untouched

Suppose the correspondence is granted in full, as a theorem about its model. What has been shown? Three things, of increasing boldness. First, a coherence result: the Marxist vocabulary of class and exploitation can be given definitions under which the two classifications provably coincide, so the tradition’s habitual pairing of them is not a confusion. Second, a sufficiency result: unequal endowments suffice to generate the paired classifications, with no assistance from coercion, authority, or any premise about production’s interior. Third, and most boldly, a suggestion about explanatory priority: since the distribution generates the pairing unaided, the distribution is the natural candidate for the phenomenon’s foundation, and the labour relation a mechanism by which the foundation expresses itself. The first two results are as solid as their model’s premises. The third is an inference, and inferences from sufficiency to priority are contestable wherever they appear in science or philosophy. Oxygen suffices, with fuel and heat, for fire; it does not follow that fire is fundamentally about oxygen rather than about oxidation considered whole. Roemer’s writings move between the second and third claims with a confidence his critics have not always shared, and a careful reader should mark the seam where proof ends and interpretation begins.

What the correspondence leaves untouched is equally worth listing, because the untouched territory is where most human experience of work resides. The model is silent about effort: how hard people work inside the contracts they sign, and how that intensity is elicited. It is silent about authority: who decides what is produced, by what methods, at what pace, subject to what appeal. It is silent about the texture of subordination, the daily experience of taking instructions from another person whose interests diverge from one’s own. It is silent about insecurity, about the difference between a livelihood that continues at another’s pleasure and one that does not. Defenders of the classical account regard these silences not as gaps awaiting a supplement but as the phenomenon itself escaping through the model’s seams. A theory of exploitation that cannot distinguish a self-employed artisan’s bad year from a sweated worker’s ordinary week, except by consulting their asset positions, has in their view mistaken the shadow for the figure. Roemer’s defenders reply that shadows are cast by figures, that the asset positions explain why the sweated worker is sweated and the artisan is not, and that a theory of the shadow’s geometry is a theory of the figure’s placement. The metaphor war is enjoyable and inconclusive. The analytical point beneath it is that explanation has levels, and the two accounts are optimised for different levels: one for the mechanism inside the relation, the other for the distribution that populates the relation’s poles.

There is also a temporal silence. The correspondence is a result about equilibrium, about where agents land once prices have settled and choices have been made. Economies, however, are trajectories, and the classical account is at its most distinctive when it describes dynamics: accumulation altering the asset distribution that generates class, class altering the political conditions under which accumulation proceeds. Roemer’s framework can be dynamised, and his later work on growth and distribution does dynamise parts of it, but the canonical statements of the criterion and the correspondence are comparative statics, comparisons between a world and a counterfactual world at a moment. Critics who charge the account with freezing history are pointing at a real feature. Defenders respond that a well-posed static question is not overturned by the existence of dynamic ones, and that the classical dynamics themselves presuppose an answer to the static question of what exploitation is at each moment along the path. The exchange is a draw at the level of method, which is why the article’s verdict, when it comes, will rest on the domination question rather than on the statics.

One more feature of the correspondence deserves a student’s attention, because it is the feature examiners reward and summaries omit. The principle is a correspondence, not an identity of definitions. Class is defined by behaviour in the labour market, what an agent finds it optimal to do. Exploitation is defined by the withdrawal counterfactual, what would happen to a coalition under an equalising departure. The two definitions are built from different materials, conduct and counterfactual welfare, and the theorem is that they sort the population identically. Results of this shape, where independently defined classifications coincide, are the strongest kind of evidence that a concept pair tracks something real in the structure being modelled, which is why the principle impressed readers across the spectrum of sympathy. It is also why the generalised theorem, to which the article turns next, cut so deep: the same style of result, pursued one step further, dissolved the privilege of the very commodity around which the entire tradition had organised its account.

The generalised commodity exploitation theorem: the centre of the argument

State the classical claim in its most defensible form. In a capitalist economy, profits are positive only if the labour workers perform outruns the labour embodied in the goods their wages command. This is the fundamental Marxian theorem, and within its assumptions it is true; Morishima’s proof is a landmark of twentieth-century Marxian economics. The claim’s defenders drew from it a moral conclusion: labour is the source of the surplus, workers are the exploited factor, and the exploitation of labour is the specific wrong at capitalism’s heart. Roemer’s generalised theorem examines the inference from the correspondence to the moral conclusion, and it does so by a manoeuvre of exemplary simplicity. Take the proof of the fundamental theorem and run it again with a different commodity in the place labour occupies. Choose steel. Define the steel value of every commodity as the total steel, direct and indirect, embodied in its production, just as labour value was defined by embodied labour. Define surplus steel analogously. Then the same formal steps establish that positive profit requires the economy to yield more steel than is embodied in the steel industry’s inputs, that steel counts, in the constructed sense, as exploited. The theorem is general: any basic commodity can be substituted, energy, fertiliser, machine tools, and the exploitation of that commodity follows with the same necessity and the same emptiness.

What does the fundamental Marxian theorem establish on its own terms?

A correspondence between two magnitudes. Under its assumptions, aggregate profit is positive exactly when the labour workers perform exceeds the labour embodied in the bundle their wages buy. It certifies that profit and surplus labour coexist, and it does not, without further argument, certify that labour is the morally distinctive source of the surplus.

Sit with the steel result until its strangeness resolves into its point. Nobody believes steel is wronged. Steel has no welfare, no claims, no standpoint from which an arrangement can be unjust to it. Yet the formal structure that was supposed to certify labour’s special status certifies steel’s with equal facility, which means the formal structure never certified special status at all. What the fundamental theorem establishes is a property of the economy’s technology and price system, a duality between profit and the surplus of any basic input over its own reproduction requirements. Labour shares that property with every other basic input because the property belongs to input-output systems as such, not to labour’s human character. The moral weight of the classical account must therefore be carried by something outside the theorem: by the fact that labour is performed by people, that its performance is their lives being spent, that the surplus is extracted from beings with claims. Roemer’s point is not that this further material is unavailable. His point is that once it is supplied, it does the argumentative work alone, and the value-theoretic machinery becomes an elaborate detour to a conclusion that a theory of just distribution reaches directly. If what makes exploitation wrong is that people end up with less than a just distribution would give them, because the asset division is unjust, then say that, and let the labour theory of value retire.

Which commodity does the generalised theorem substitute first in this article?

Steel, following the illustration Roemer’s own presentations made standard. Steel values are defined by the steel embodied directly and indirectly in each commodity, surplus steel is defined in parallel with surplus labour, and the identical formal argument then certifies that positive profit requires the exploitation of steel, which nobody believes is wronged.

The theorem’s reception divided along a line that predicts much of the subsequent debate. Economists, including many with no Marxist sympathies, received it as a clean and decisive result about what value theory can and cannot show, and it is the part of Roemer’s work most often cited in technical literature. Marxist philosophers and political theorists were more likely to respond that the theorem refutes a claim nobody careful had made. The classical account, on this response, never rested labour’s specialness on the formal correspondence alone. It rested on the analysis of the labour process, on the distinction between labour and labour power, on the fact that the worker accompanies her commodity into production and experiences its consumption from inside. Steel does not accompany anything. The theorem, the response concludes, demolishes a formalist misreading of Marx and leaves the substantive account standing. Roemer’s counter-response is the relocation itself: strip the substantive account of the value machinery, and what remains standing is a claim about the distribution of assets and the relations it generates, which is his account, stated in the tradition’s older vocabulary. The dialectic at this point becomes genuinely difficult to referee, because both sides claim the same residue as their own. The article’s verdict section will propose a way of scoring the dispute, and will attach to its score the defeater the brief requires: a statement of what would reverse it.

Why does the substitution leave the formal argument intact?

Because the proof uses only properties that every basic commodity shares with labour: it is used up in production, it is embodied indirectly in other goods through the input structure, and its reproduction requires a definite quantity of itself. The algebra of input-output systems is indifferent to which input is singled out for the accounting.

Before the objections are heard, the consequence bearing on the labour theory of value should be stated with precision, since the slogan version of it, that Roemer disproved the theory, is wrong and the article corrects misreadings as it goes. The labour theory of value, as a theory of prices, faces difficulties that belong to its own article in this series and are not relitigated here. The generalised theorem addresses a different question: whether the theory is needed to ground the exploitation claim. Roemer’s answer is that it is not, and the answer has two parts. Negatively, the theorem shows the value accounting cannot privilege labour. Positively, the withdrawal criterion supplies a ground for exploitation claims that never mentions embodied labour, so the ethical argument survives the value theory’s removal intact, indeed survives more cleanly, since it no longer depends on contested price theory. Roemer drew the conclusion about his own intellectual allegiances explicitly: the labour theory of value should be abandoned as a foundation, and Marxists should rebuild their critique of capitalism as a critique of unjust asset distribution. That a leading figure of the analytical school said this in print, as a conclusion of proofs rather than as a polemical flourish, is what makes the episode a turning point in the tradition’s modern history rather than a footnote to it.

The consequence: the ethical argument changes its address

Follow the relocation to its end and the critique of capitalism changes its address, from the factory to the constitution of ownership. On the classical account, a capitalist economy is objectionable because of what its central transaction does: the wage contract, fairly struck and honestly honoured, still delivers the worker’s surplus labour to the employer. On Roemer’s account, the same economy is objectionable because of what its participants bring to the transaction: a distribution of productive assets so unequal that one class must sell its labour to another on pain of destitution. The transactions themselves, considered one by one, may be mutually beneficial, voluntarily accepted, and preferable to every alternative the parties actually face. The wrong sits behind them, in the endowments. This is why Roemer’s framework belongs as much to the theory of distributive justice as to Marxian economics, and why its natural interlocutors became philosophers of justice, egalitarians arguing about what an equal distribution is a distribution of, rather than only economists arguing about value. The framework also explains its author’s trajectory. A theorist who locates the wrong in the starting distribution will eventually be asked what starting distribution would be right, and what institutions could establish and protect it, and Roemer’s later career, the turn to market socialism and to equality of opportunity, is that question pursued across three decades.

What follows for the ethical argument if labour is not special?

The objection to capitalism must be re-founded. If the formal argument privileges no factor, then the wrong of exploitation cannot rest on labour’s unique role in production, and must rest instead on the injustice of the asset distribution that makes one class dependent on another, which is a claim in distributive justice.

Notice how the re-founding redistributes argumentative burdens. The classical defender had to win on value theory before the ethical conversation could begin, and generations of debate were consumed at that gate. The Roemerian critic of capitalism concedes the gate is unnecessary and walks around it, but arrives at a different gate: the theory of just acquisition and transfer. Is the existing asset distribution unjust, and by what standard? Standards compete. A libertarian standard, under which holdings arising from just acquisition and voluntary transfer are just whatever their pattern, will find no exploitation in Roemer’s sense even where the withdrawal test is satisfied, because the test measures gain from equal division and a libertarian denies that equal division is the benchmark of justice. An egalitarian standard finds the test’s benchmark natural. Roemer was aware that the criterion presupposes a benchmark, and his defence of the egalitarian benchmark belongs to his work on distributive ethics rather than to the theory of exploitation proper. The honest summary is that the withdrawal criterion does not settle the justice of distributions; it shows what follows, for the classification of positions, once an egalitarian benchmark is adopted. Critics who present the criterion as smuggling egalitarianism into a definition have hold of something true, and defenders who present it as deriving egalitarian conclusions from neutral premises overclaim. The criterion is a bridge: distributive ethics on one side, class analysis on the other, with traffic flowing both ways.

Which premise about production does the new account simply not contain?

Any premise at all about what occurs inside it. The account neither asserts nor denies that workers are directed, that effort is extracted, or that the working day divides into necessary and surplus portions. Production enters as a technology that transforms inputs into outputs, and its interior organisation is outside the model’s field of vision.

That absence is the hinge on which the whole assessment swings, so it should be examined from both sides before the named objections are introduced. From the account’s side, the absence is methodological hygiene. A definition of exploitation that presupposes a theory of the labour process cannot be used to test claims across social forms where labour processes differ radically, whereas a definition built on endowments travels: the same instrument pronounced on feudal dues, capitalist wages, socialist skill premiums, and bureaucratic privilege in the taxonomy set out earlier. Generality is a theoretical virtue, and the price of generality is always abstraction from local texture. From the critics’ side, the absence is the whole story. The labour process is not local texture. It is where human beings spend their waking lives under the direction of others, and a concept of exploitation purified of it has been purified of the experience that made the concept worth having. Both sides are describing the same absence accurately. They value it oppositely, and the valuation cannot be settled by further definition. It can only be settled, if at all, by asking what work the concept is wanted for, and that question is political as much as analytical. The verdict section returns to it with a proposal.

The Marxist objections: domination, compulsion, and the emptied concept

The objections now to be heard are not the complaints of readers who found the mathematics uncongenial. They are the considered replies of theorists inside or near the tradition, they were stated in print with arguments, and the brief for this article requires them at full strength with attribution. Three objections organise the critical literature. The first charges that the relocation loses domination. The second charges that the framework builds voluntariness into the exchange it analyses, assuming away the compulsion that propertylessness constitutes. The third charges that the criterion’s verdicts, in cases its author concedes, detach the concept from the judgements it was coined to express. Each objection will be stated, sourced, and then answered as Roemer answered it or as his framework permits, with the concessions the answers require made explicitly.

The domination objection, in its most developed form, is associated with the philosopher Nicholas Vrousalis, whose work on exploitation and domination argues that the phenomenon Marx named is at its core a relation in which one party exercises power over another’s labour for the dominator’s enrichment. On this view, enrichment through another’s service is not sufficient for exploitation, and neither is an unjust background distribution. What is necessary is the relational exercise of power: the employer does not merely benefit from the worker’s asset poverty but directs the worker’s activity, hour by hour, in a process the employer controls. Vrousalis’s argument proceeds partly by cases. Imagine a distribution of assets that is unequal, generating transfers from asset-poor to asset-rich through market exchange, but in which no one directs anyone’s labour, as in Roemer’s own credit-market construction, where lenders benefit from borrowers without supervising them. The domination account denies that the borrowers are exploited in the sense that matters, however the distribution is judged. Conversely, imagine a workplace in which a manager directs workers’ labour for an organisation whose assets are collectively owned, so that no asset inequality explains the relation. The domination account can still find exploitation there, in the directed extraction of effort, where the withdrawal criterion, applied to the asset test, finds none. The cases are constructed to pry the two accounts apart, and they succeed in showing that the accounts are not two descriptions of one phenomenon but two phenomena sharing a name.

How does the domination objection describe the supervised worker?

As the paradigm the concept exists to name. A person whose activity is directed by another, whose effort is monitored and paced, and whose product is appropriated by the directing party, is exploited in virtue of that relation, and an account that classifies her by her asset position alone has described her circumstances while omitting her subjection.

G. A. Cohen’s work supplies the second objection’s sharpest instrument, though Cohen aimed it at a neighbouring target. In his analysis of proletarian unfreedom, Cohen argued that workers are collectively unfree to escape the proletarian condition even when each individual worker faces a genuine choice among employers, because the class’s exit is blocked by the structure of ownership: if all workers attempted to leave wage labour, the assets available to sustain them outside it would not stretch, so the freedom each enjoys is a freedom to change masters within a condition none can leave. Roemer knew this argument intimately, since Cohen was a fellow founder of the analytical school and the two positions developed in dialogue. The objection as pressed against Roemer runs as follows. The model’s agents choose optimally given their endowments, and the framework’s language of choice, preference, and voluntary contract systematically represents as free a situation whose essence, on the classical view, is compulsion by circumstance. Propertylessness is not one endowment among others, like a small bank balance. It is the condition of having nothing to sell but one’s capacity to work, and a framework that registers this condition only as a low number in an asset column has represented a social relation as a quantity. Roemer’s reply concedes the representation and disputes the loss: the quantity, he argues, is what does the explanatory work, since varying the number varies the outcome, and the language of compulsion, however evocative, names the same dependence the model measures. The reply is characteristic of the whole encounter between the analytical school and its critics, summarised in the series article on the criticisms of analytical Marxism: one side’s evocative naming is the other side’s unexamined metaphor, and the argument is over which failure is worse.

The third objection, concerning counterintuitive verdicts, has already been partly heard in the section on the criterion’s awkward cases, and needs only its theoretical moral drawn. A definition is a tool for sorting, and tools are judged by whether their sorting serves the purposes for which sorting is wanted. The withdrawal criterion sorts some comfortable people into the exploited class and some wretched people out of it. If the purpose of the concept is to identify whose disadvantage a just redistribution would remedy, the sorting is defensible, even illuminating. If the purpose is to identify the relations of power that a transformation of production would have to abolish, the sorting misleads. Roemer’s own writings acknowledge the criterion’s verdicts without flinching, including the verdict about the talented under socialism, and treat the flinching of others as evidence that they want the concept to do work it cannot do. The standoff is instructive because it is not resolvable by closer attention to Marx’s texts. Both purposes are genuinely present in the tradition. The texts support a relational reading and a distributional reading in different passages, which is one reason the tradition’s heirs can each claim fidelity while accusing the other of abandonment.

The strongest objection, stated at full strength and answered

The brief governing this article requires the strongest objection to be stated as its holders state it, without softening, and then answered, with the answer’s concessions named. Here is the objection in that form. Exploitation, in Marx, is a relation in which one party’s labour is directed by another and its product appropriated by that other. The relation is the phenomenon. An account that reaches identical verdicts from the distribution of assets alone, without anywhere representing the direction or the appropriation, has described a symptom and omitted the disease. Worse, the omission is not neutral between cases the classical account distinguishes with ease. A worker labouring under close supervision, her pace set by a line she does not control, her product claimed by the firm that directs her, and a small asset-holder who employs no one and is employed by no one but rents a modest property to a tenant, can emerge from the withdrawal test with the same exploitation status, or with their statuses reversed, depending on their asset positions. The classical account exists precisely to deny that these two people stand in the same relation to the economy’s wrongs. A concept that cannot mark the difference between being directed and merely being endowed has not been generalised. It has been evacuated.

The answer begins with the correspondence principle, which is why that principle occupies so much of this article. The principle shows that, under the conditions of a capitalist economy as modelled, the people who end up directed and the people who end up directing are sorted into those positions by their endowments, reliably and without exception inside the model. The supervised worker is supervised because her asset position made selling her labour her best option, and the supervisor directs because his asset position made hiring his best option. The relation the objection prizes is therefore not independent of the distribution the criterion measures. It is the distribution’s predictable product, and a theory of the product’s distribution is a theory of why the relation’s poles are occupied by the people who occupy them. When the objection says the account describes a symptom, the account replies that it describes the cause of the symptom’s incidence: not why supervision exists as an institution, but why this person is supervised and that person supervises, which is the question an indictment of particular arrangements must answer. Add that the criterion’s defenders never claimed the interior of the labour process is unimportant for every purpose. They claimed it is unnecessary for the exploitation verdict, and the correspondence is their evidence.

Then the concession, which the brief requires and honesty would require anyway. The framework holds no premise about production at all, and therefore cannot distinguish domination from inequality. Where direction and appropriation occur without the relevant asset inequality behind them, the criterion is blind, and where asset inequality occurs without direction, the criterion pronounces a verdict the classical ear hears as a category error. Roemer’s framework does not rebut this. It absorbs it, by denying that the category error is an error: on the relocated concept, the lender exploiting the borrower through interest, with no supervision anywhere in sight, is not a metaphorical extension of exploitation but a central case of it, and the worker supervised in a collectively owned firm is a victim of domination, a genuine and serious wrong, which is simply a different wrong from exploitation. The vocabulary is being deliberately redrawn, with exploitation assigned to the distributional axis and domination to the relational axis, and the redrawn map is offered as clearer than the inherited one, in which a single word carried both axes and confused arguments that needed them separated. Whether the redrawing clarifies or amputates is the judgement the reader is now in a position to make, and the verdict section below makes it with its defeater attached.

How does Roemer answer the charge that the concept has been emptied?

By pointing to the correspondence. The relations the critics prize are populated by the distribution the criterion measures, so the account explains who directs and who is directed even though it never describes directing. What the critics call the phenomenon, he treats as the distribution’s expression inside production.

What would defeat the judgement that domination is lost?

A formulation of the withdrawal criterion that separated exploitation from mere asset inequality by reference to a relation obtaining during production would defeat it. Roemer’s framework contains no such formulation, and this article states the defeater openly and invites the construction it describes.

Pause over the defeater’s shape, because it disciplines the verdict. The judgement that the account loses domination is a judgement about a specific framework, the one Roemer built, not about every possible descendant of it. A successor theory that kept the withdrawal test’s distributional core while adding a production-relation condition, exploited only if the coalition would gain from withdrawal and its members’ labour is directed by others, would no longer lose domination, and the judgement recorded here would fall. Nothing in the present framework supplies that condition, and bolting it on would cost the framework its proudest results: the credit-market isomorphism would fail, since lending involves no direction, and the taxonomy’s generality would narrow to economies with workplaces. The defeater is therefore not a loophole but a price tag. It tells the reader exactly what saving domination inside this research programme would cost, and lets the reader decide whether the price is worth paying. Verdicts in this series are required to carry their defeaters in this way, and this article’s central verdict is no exception: the relocation succeeds as an account of exploitation’s distributional basis and fails as an account of exploitation’s relational reality, and the failure half of that sentence stands only until a formulation meeting the stated condition is produced.

There is a further reply available to Roemer’s defenders that deserves hearing because it turns the objection’s own example. Take the small asset-holder with the tenant, whom the objection uses as the criterion’s embarrassment. The classical account, focused on production, has its own embarrassment with this figure: rent extraction from a tenant involves no labour process at all, yet the tradition has always wanted a vocabulary for the transfer involved, and has reached for exploitation-adjacent terms, usury, rentier parasitism, unearned income, without a unified account of what makes the transfer wrongful. The withdrawal criterion unifies the cases the tradition handled piecemeal. Wage exploitation, interest exploitation, and rent extraction become one phenomenon, differential command over resources rooted in differential endowment, expressed through different contracts. Unification is a genuine theoretical achievement, and the history of science honours it in other fields. The objection’s force is not that unification is worthless but that this unification is purchased by discarding a distinction, between transfers backed by direction of persons and transfers backed only by ownership of things, that moral judgement may need. Both the achievement and the price are real. An assessment that mentions only one of them has not finished its work.

What the relocation gains: tractability, generality, and a usable indictment

Justice to the account requires a section on its gains stated as warmly as its defenders state them, since the steelman discipline of this series applies to the article’s subject and not only to its subject’s critics. The first gain is tractability. Exploitation claims become checkable. Given an ownership distribution and a model of the economy, the criterion returns verdicts by computation, and disagreements about the verdicts can be traced to disagreements about the distribution data or the model, which are the kinds of disagreements that evidence can settle. Compare the classical claim’s evidentiary demands, which require agreement on value magnitudes that are not directly observable and whose measurement has consumed a literature. The second gain is generality, already displayed in the taxonomy: one instrument, four social forms, and a principled way of saying what successive emancipations leave unresolved. The third gain is the one Roemer himself emphasised most in his later reflections. The relocated concept supports an indictment of capitalism that does not require the indictor to defend the labour theory of value, the transformation of values into prices, or any particular reading of the classical texts. The indictment runs: the distribution of productive assets is unjust, positions generated by an unjust distribution inherit its injustice, and the inheritance is measurable. A critic of capitalism who finds the classical foundations contested can hold this indictment without embarrassment, and a defender of capitalism knows exactly which premise to attack, the justice of the distribution, rather than shadow-boxing with value theory.

How did economists outside Marxism take up the formalisation?

As a contribution to the theory of inequality and property rights. The models of agents sorting by endowment, and the results connecting asset distributions to class-like structures, entered general economic discussion detached from their author’s allegiance, and the withdrawal idea was taken up in political philosophy’s debates about justice in holdings.

The gains have a common root, and naming it clarifies the losses as well. Roemer’s account treats exploitation as a property of distributions of a transferable stock. Everything transferable can be counted, compared, counterfactually redistributed in imagination, and modelled. The account’s tractability and generality are the virtues of countability. Domination is not countable in that way; it is a quality of a relation between persons, resistant to aggregation, varying with the texture of authority rather than with the size of endowments. A theory built on a countable stock will therefore always find domination slipping through its mesh, not because its author was careless but because the mesh was woven for a different catch. The mature assessment of the relocation is that it is a triumph of one theoretical style, the economist’s, applied to a concept that also belongs to another style, the critic of relations, and that the concept’s subsequent history in the literature is the story of the two styles negotiating a border. Vrousalis’s domination account is one border proposal, from the relational side. Roemer’s market socialism, to which the article turns now, is a proposal from the distributional side about what to build once the diagnosis is accepted.

The market socialist turn: from defining exploitation to designing its remedy

A theorist who concludes that exploitation is unjust asset distribution wearing institutional clothes faces an obvious next question. What distribution, secured by what institutions, would make the clothes innocent? Roemer’s answer, developed most fully in A Future for Socialism in 1994, is the destination toward which the theory of exploitation had been travelling since 1982, and it cannot be understood apart from the diagnosis it remedies. If the wrong is that a small class owns the economy’s productive stock while the majority own almost none, the remedy is not necessarily to abolish markets, firms, or profit-seeking managers. The remedy is to abolish the concentrated ownership: to distribute the claims on the economy’s capital so widely that the class structure the correspondence principle describes cannot form. Market socialism, in Roemer’s design, keeps the market’s informational and motivational machinery and socialises the ownership of the firms that machinery coordinates. The design question, whether planning or markets should coordinate production, is treated in this series at can planning beat the market, and this article touches the design only where Roemer’s specific mechanism, the coupon, expresses the theory of exploitation in institutional form.

What problem was coupon socialism designed to solve?

Reconcentration. Any scheme that distributes ownership equally at a stroke faces the question of what stops the equal shares from being sold, accumulated, and reconcentrated into a new owning class within a generation. The coupon mechanism answers by making the distributed claims non-transferable for cash, so equality of ownership is a permanent constitution rather than a founding moment.

The mechanism, in outline, runs as follows. Every adult citizen receives an equal allotment of coupons, a parallel currency usable for one purpose only: purchasing shares in the economy’s firms, held through mutual funds that citizens choose among. Coupons cannot be sold for ordinary money, cannot be converted into consumption goods, and cannot be accumulated by purchase from other citizens, because no market in coupons exists. Shares bought with coupons yield dividends, which are ordinary income and can be spent as the citizen wishes. When a citizen dies, the coupons and the share portfolio revert to the public stock, to be redistributed to the next cohort reaching adulthood, so the equality of the starting allotment is renewed each generation rather than inherited away. Firms, owned through the mutual funds, are managed by professionals whose performance the funds monitor, and the funds, competing for citizens’ coupons, have incentives to monitor seriously. Large firms in Roemer’s sketch have a public presence on their boards and access to investment funds through a banking system with public obligations, adapting features he drew from the Japanese and German corporate governance of the period. The state’s role shrinks to maintaining the constitution of ownership and the framework of competition, rather than directing production.

How do coupons prevent the reconcentration of ownership?

By blocking every channel through which equal shares become unequal. Coupons cannot be sold, so the improvident cannot alienate their stake and the acquisitive cannot buy it. Portfolios revert at death, so advantage cannot compound across generations by inheritance. What a citizen can accumulate is dividend income, not a larger share of the capital stock itself.

Assess the design against the theory that generated it, and the fit is exact to a degree rare in political economy. The withdrawal criterion defines exploitation by the gap between actual positions and positions under equal per capita asset ownership. Coupon socialism is the institutionalisation of the counterfactual: it makes the actual distribution approximate the equal division the test imagines, permanently. Under a functioning coupon constitution, no coalition would gain from the imagined withdrawal with a per capita share, because the actual share already is the per capita share, and capitalist exploitation in Roemer’s precise sense would be abolished, not by exhortation but by construction. The class-exploitation correspondence then predicts the rest. With endowments of alienable assets equalised, the sorting of agents into hirers and sellers of labour loses its distributional engine; people would still work for firms, and firms would still hire, but the positions would reflect preferences and skills rather than a compulsion rooted in asset poverty, and the wage relation would persist as a form while its exploitative content, on this account, drained away. Whether one finds that prospect inspiring or chilling depends on whether one accepts the relocation, which is why the theory and the design stand or fall together in most readers’ judgements.

The design attracted criticism from several directions, and the series discipline of steelmanning applies to it as well. Market socialists of other schools, David Schweickart most prominently, argued that Roemer’s firms, owned through funds and managed for shareholder return, replicate too much of capitalism’s internal life, including its indifference to workplace democracy, and that a socialism which changes only the share register has conceded the argument about production that the classical tradition cared most about. The criticism is the domination objection translated into institutional design, and it lands with the same force and meets the same reply: Roemer’s answer is that the share register was the argument, and that workplace relations under equal ownership would transform themselves as the bargaining positions behind them equalised. Defenders of private ownership attacked from the other side, questioning whether non-transferable coupons could sustain investment discipline and whether the reversion at death would distort saving and enterprise across the life cycle. Roemer’s responses, in the book and in the comparative systems literature that discussed it, were characteristically concessive in form: the design was a sketch of a constitution, offered to show that egalitarian ownership and market coordination are compatible in principle, not a blueprint claiming to have solved every incentive problem in advance. The sketch’s afterlife in policy discussion, where voucher and citizen-ownership schemes recur in debates far from Marxism, testifies to the design’s portability once its theoretical parentage is forgotten.

Why does Roemer keep markets in his socialism?

Because his diagnosis never indicted markets. The wrong, on his account, is the distribution of assets with which people enter markets, not the existence of exchange, prices, or firms. Reform the endowments and the market becomes an instrument that allocates without exploiting, so the design socialises ownership and leaves coordination to prices.

One further element of the later work belongs in this article because it completes the relocation’s arc. Roemer’s attention shifted, across the nineteen nineties and after, from the distribution of assets to the distribution of opportunities, and his equality of opportunity framework asks what share of a person’s outcome is attributable to circumstances beyond her control, birth, family, social environment, as against effort for which she can be held responsible. The continuity with the exploitation theory is structural. In both, the moral charge attaches to a background distribution the individual did not choose, and the remedy equalises the background while leaving the foreground of choice and reward intact. The talented person whose skill commands a premium troubled the withdrawal criterion as the problem of socialist exploitation. In the opportunity framework, the same figure reappears as the person whose circumstances included the development of a valuable talent, and the framework’s machinery of circumstance and effort is an attempt to price that figure’s claim more finely than the criterion could. Readers tracing the arc see a single intuition refined across a career: justice is about the fairness of starting positions, and the critique of capitalism is the critique of a starting gate that some reach with capital and others reach with nothing but their hands.

The books and the building of the argument

The argument arrived in stages, and the stages matter because critics sometimes attack an early formulation that its author had already revised. The foundation is A General Theory of Exploitation and Class, published in 1982, a book whose title announces its ambition: not a commentary on Marx but a general theory, covering feudal, capitalist, and socialist forms under one instrument, with class derived rather than assumed. The book’s method set the pattern for the analytical school described in this series at analytical Marxism: the complete guide: state the economy as a model, define the contested concept inside the model, prove what follows, and let the philosophy be drawn out of the proofs by hand afterwards. Its central constructions, the withdrawal definitions, the correspondence principle, and the generalised theorem, are all present in that first book, which is why the secondary literature cites 1982 as the relocation’s date.

Free to Lose, published in 1988, carried the argument to a wider audience and is the better entry point for most readers. Written as an introduction to Marxist economic philosophy, it restates the theory with less formal apparatus and more explicit attention to the ethical conclusions, and it contains Roemer’s clearest statements of why the labour theory of value can be relinquished without relinquishing the critique of capitalism. The title itself is an argument: the freedom the propertyless worker enjoys, the freedom to sell labour to any employer or none, is the freedom to lose, since declining every offer means destitution, and the phrase condenses the compulsion objection into four words while relocating its basis. Roemer does not deny the compulsion. He prices it, as the difference between the worker’s position and the position an equal asset share would give her, which is the withdrawal test doing biographical work inside a book title.

Between and around the books sit essays that sharpened particular joints of the theory. The essay asking whether Marxists should be interested in exploitation, published in the mid nineteen eighties, is the most philosophically direct statement of the relocation: it argues that the concept of exploitation is not the moral foundation of the Marxist critique but a derivative notion, useful when it tracks the underlying injustice of distributions and dispensable when it does not. That ordering, distribution first and exploitation second, inverts the tradition’s usual presentation, in which exploitation is the bedrock wrong and distributive claims are its consequences. Readers who want the relocation defended as philosophy rather than as economics should start there. The later work then changes register. The 1994 volume on socialism’s future is institutional design. The writings on equality of opportunity, developed across the following decade, are distributive ethics with formal apparatus. The exploitation theory is the root system of all of it, and the root is easiest to see in the 1982 book, before the branches thickened.

Why is the founding book called a general theory?

Because its instrument is not built for capitalism alone. The withdrawal definitions are parameterised by what the departing group carries and by which assets are imagined equalised, so one framework generates accounts of feudal, capitalist, and skill-based exploitation, and generality across social forms is the property the title claims for the construction.

A word about the book’s difficulty is owed to students approaching it. The formal core is genuinely formal, written in the idiom of mathematical economics of its period, and readers without that training will find the proofs demanding. The philosophical yield of the proofs, however, is recoverable from the statements of the results and from the author’s own summaries, and Free to Lose exists precisely to make that recovery possible. Seminar leaders exploit this two-tier structure: the criterion and the correspondence can be taught, debated, and examined from the accessible statements, while the proofs stand behind the statements for those equipped to audit them. The arrangement has a cost that critics noted early. A theory whose authority rests on proofs most of its audience cannot check asks for a kind of trust that sits oddly with its own rhetoric of transparency, and the analytical school’s reputation for technocratic enclosure owes something to this structure. The defence is that the statements are checkable in the only way that finally matters: their consequences can be traced, their counterexamples constructed, and their premises debated, all without reproducing the proofs, and this article has proceeded on exactly that basis.

The reception history completes the picture of the books. Within Marxian economics, the 1982 book was recognised immediately as a major event, reviewed extensively, and argued with for a decade. Within mainstream economics, it was received as an interesting formalisation by a theorist whose other work, on general equilibrium and later on political economy, commanded independent respect, and the exploitation models entered citation networks that had never previously carried Marxian results. Within political philosophy, the withdrawal criterion and the property-relations account of exploitation became standard reference points in the literature on justice in holdings, cited by egalitarians and by their libertarian critics alike. Few books in the Marxist tradition have been read by so many readers who reject the tradition, and the portability across hostile audiences is itself evidence about the relocation: a concept rebuilt on distribution theory’s foundations can travel through distribution theory’s republic of letters, carrying a Marxian passport that is inspected, stamped, and argued over at every border.

A small economy, examined twice

Abstractions in this territory repay a concrete walk-through, and the walk-through can be conducted entirely in words, with a stipulated little economy whose features are chosen for clarity rather than realism. Imagine a community of households producing cloth and grain with looms, ploughs, and land. In the first version of the story, the instruments are held in common and each household works them in turn, keeping what it produces and trading surpluses with neighbours. In the second version, the same instruments, identical in number and quality, are owned by a tenth of the households, and the remaining nine tenths own nothing but their capacity to work. The second version’s households must therefore make terms with the owners: they work the owners’ looms and ploughs and receive a portion of the product, while the owners, who also work, receive their own product plus the retained portion from every household they accommodate. Nothing else differs between the versions. The technology is the same, the people’s skills and preferences are the same, the weather is the same.

Run the classical account over the second version and it reports as follows. The labouring households work longer than the labour represented in the portion they receive, since the owners’ retained portion embodies labour the owners did not perform. Surplus labour is extracted inside production, exploitation is present, and its measure is the ratio of retained to received portions. Run the withdrawal criterion over the same version and a parallel report comes back. Take the coalition of the nine tenths. Imagine it departing with nine tenths of the looms, ploughs, and land, a per capita share being exactly that. Working its own instruments and keeping its whole product, the coalition is better off than it is under the owners’ terms, so the coalition is exploited, and the owners, who would lose by an equalising departure, are exploiters. The two accounts agree on the verdict, as the correspondence principle says they should in economies of this shape, and the agreement is worth pausing over, because polemics on both sides sometimes proceed as if the accounts were rival verdict machines returning opposite answers. In the central cases, they return the same answers. The dispute is about what makes the answers true.

Now vary the story in the direction the theorem explores. Keep the second version’s ownership exactly as it is, but suppose the community’s weaving technology changes so that cloth production depends crucially on a scarce dye-stuff, gathered and processed by methods the owners also control. The classical account’s verdict about the labouring households is unchanged: surplus labour is still extracted, in whatever quantity the new technology implies. The generalised theorem now invites a parallel computation for the dye-stuff. Define the dye embodied in each unit of cloth, directly and through the dye’s own production requirements, and the same formal pattern emerges: the economy’s surplus product exists only because more dye is gathered than is embodied in the dye-gatherers’ consumption, and the dye emerges, in the constructed sense, as exploited. The parallel is exact and the conclusion is absurd, and the absurdity is the argument. Whatever entitles us to say the labourers are exploited, it cannot be the formal pattern, since the pattern certifies the dye as well. It must be something about labourers that dye lacks, and the candidates are all facts about persons: that labourers are beings with welfare and claims, that their labour is their life-activity, that the distribution of the instruments they must use is a human arrangement open to moral assessment. Roemer’s relocation consists of building the definition directly on those facts and discarding the formal pattern as a foundation, while retaining it as a diagnostic of how economies hang together.

Vary the story once more, in the direction the domination objection explores. Return to the first version, instruments held in common, but suppose the community appoints overseers to coordinate the work in the fields and sheds, and suppose the overseers grow accustomed to command: they set the pace, assign the tasks, punish slackness by repute and by roster, and enjoy deference and the best portions at the common table, though the instruments remain everyone’s and the product is shared by the standing rules. The classical account has resources here that commentators dispute, since no surplus is privately appropriated, yet direction and subjection are plainly present. The withdrawal criterion, applied to alienable assets, finds nothing: no coalition gains from departing with instruments, divided per head, that it already shares. Applied to status, in the taxonomy’s fourth cell, it can register the overseers’ privilege, since a coalition might gain from a departure in which offices carried no perquisites. The little economy thus exhibits, in three variations, the whole argumentative geography of this article: agreement in the central case, the theorem’s unseating of labour’s formal privilege, and the relational residue that the asset test cannot see until its status extension is invoked, an extension whose adequacy the reader must judge.

Why does the little economy produce the same verdict twice?

Because its shape is the shape the correspondence principle describes. Ownership is concentrated, the asset-poor must work the owners’ instruments to live, and the retained portion is the visible form of the gain the owners would lose under an equalising departure. Where class and endowment align so cleanly, the two accounts are two descriptions of one structure.

The walk-through also exposes a question the criterion’s statement leaves open and the little economy makes vivid. When the nine tenths depart with their per capita share, what happens to the remaining tenth is left out of the computation, yet the owners’ loss is the mirror of the coalition’s gain only if the total product survives the partition. In the story, production is simple enough that partition costs little. In economies where production depends on large indivisible installations, a railway network, a power grid, a port, the per capita share of such assets is not a thing a departing group can carry in any operationally meaningful sense, and the counterfactual’s coherence strains. Roemer’s constructions handle indivisibility by treating shares as claims rather than as physical portions, a departing group carrying its proportional claim on the indivisible stock’s services. The repair is legitimate within the model’s logic. It also moves the test further from anything imaginable as an event and deeper into the territory of pure accounting, which is where its critics always said it lived. A test that began as a vivid thought experiment, a group walking away with its fair share, ends as an accounting identity over claims. The vividness was pedagogically valuable and theoretically dispensable, and Roemer’s own presentations move between the two registers without always marking the border.

Inside the school: agreement, friction, and the division of labour

Roemer’s reconstruction was not a solitary enterprise, and its meaning is sharpened by the company it kept. The analytical Marxism of the nineteen eighties, the September Group and its penumbra, divided the tradition’s inheritance among its members with something like a research programme’s coordination. G. A. Cohen took the philosophy: the defence of historical materialism’s functional explanations, then the long engagement with distributive justice that produced his critique of Rawls and of the incentives argument. Jon Elster took the methodology and the theory of action, pressing rational choice reconstruction on Marx’s claims about class formation and ideology, and proving a sterner critic of loose functional explanation than anyone outside the school. Erik Olin Wright took the empirical sociology, building class maps that survey research could populate. Roemer took the economics. The exploitation theory was his assigned province in this division, though assignment is too strong a word for what was really a convergence of temperaments, and the theory bears the programme’s marks throughout: clarity about premises, suspicion of dialectical shortcuts, willingness to follow a reconstruction past the point where the reconstructed doctrine changes character.

The frictions inside the school are more instructive than its agreements. Cohen’s defence of the labour theory of value’s conceptual core, in his 1979 essay on the concept of exploitation, argued that the theory’s point was never primarily price prediction but the identification of the worker as the producer of the product’s value, and hence of the capitalist’s receipt as the receipt of something created by another. That identification, producer and product, is a relational and almost juridical idea: the worker makes the thing, the capitalist takes the thing, and the taking is the wrong. Roemer’s theorem does not touch that idea directly, since the idea does not rest on the formal correspondence the theorem dissolves. But the relocation renders the idea otiose for the exploitation verdict, and Cohen’s later work on justice can be read as absorbing the blow by relocating his own centre of gravity, from exploitation to equality, from the wrong of the taking to the injustice of the distribution that makes the taking possible. The two founders thus converged from opposite directions on the territory Roemer had staked: by the nineteen nineties, the school’s most influential voices agreed that distributive justice was where the critique of capitalism had to be fought out, and disagreed mainly about the currency and pattern of the just distribution. Historians of the school sometimes present this convergence as Marxism dissolving into liberal egalitarianism. The participants presented it as Marxism’s questions surviving its answers.

Elster’s friction with Roemer ran along a different seam. Where Roemer modelled agents as optimising given endowments and preferences, Elster’s work on preference formation asked where the preferences came from, and his account of adaptive preferences, the sour grapes mechanism by which the deprived cease to want what they cannot have, poses a problem for any welfarist test. The withdrawal criterion judges a coalition’s position by the coalition’s own valuation, as an earlier section noted. If deprivation has already deformed the valuation, schooling the asset-poor to modest expectations, the criterion’s readings are contaminated at the source: a coalition that would not gain by its own lights may fail to gain partly because its lights have been dimmed by the very arrangement under assessment. Roemer’s framework can respond by idealising the preferences, evaluating the counterfactual under the preferences the coalition would have absent the deformation, but each idealisation moves the test further from computation and closer to the moralised judgement it was designed to replace. The friction is recorded here without a winner, because the literature records none: it marks a genuine frontier where the analytical programme’s tools, optimisation and counterfactual welfare comparison, meet a phenomenon, preference endogeneity, that the tools handle awkwardly wherever they are used, in Marxian applications and in welfare economics generally.

What did Cohen’s producer argument preserve that the theorem dissolves?

The identification of the worker as the maker of the product and of the capitalist as the receiver of what another made. Cohen’s early defence located the wrong in that making and taking, a relational fact about persons and things, which survives the theorem because it never depended on the formal pairing of profit with surplus labour.

Wright’s engagement, already noted in connection with class analysis, adds the empirical dimension the other founders lacked. Survey-based class research found the population distributed across locations that a single asset dimension could not generate: managers commanding organisational assets they did not own, credentialed experts commanding scarce skills, small employers suspended between the great classes. Wright’s response was to pluralise the asset base of class theory, treating organisational position and credentials as further dimensions of exploitation-relevant advantage, a move he explicitly framed as generalising Roemer’s property-relations approach rather than abandoning it. The generalisation has a price that Wright acknowledged. Each added dimension multiplies the class map’s cells and dilutes the correspondence’s clean coincidence of class and exploitation, until the map’s usefulness becomes an empirical question about which cells are populated rather than a theorem about which must coincide. The trajectory from Roemer’s parsimony to Wright’s pluralism is the standard trajectory of a formal result meeting data, and it cuts both ways in assessment: the parsimony was the proof’s enabling condition, and the pluralism is the proof’s fate in a world that refuses to be as simple as its models.

Outside the school, the property-relations account entered political philosophy’s mainstream treatment of exploitation, where it now sits in textbooks beside the classical account and the domination account as one of the three standard options a theorist must position against. Its presence in that company changed the questions syllabuses ask. Where an older syllabus asked students to expound and assess Marx’s theory of exploitation, the newer one asks which of the rival accounts captures the phenomenon, a comparative framing in which Roemer’s relocation is no longer a heresy inside a tradition but a candidate in an open competition. Competitions of that kind are judged by the criteria this article has used throughout: what the account locates, what it explains, what it implies, and what it cannot capture, the four rows of the comparison table above the verdicts. It is a mark of the relocation’s success that the competition’s very format, the matched comparison of locations and residues, is one the relocated account’s clarity made natural.

Misreadings to correct

Five misreadings circulate wherever this material is summarised, and the article corrects them in one place so that the verdict can proceed on an accurate record. The first misreading holds that Roemer’s account denies that exploitation exists. It does not. It defines exploitation, derives its incidence, and condemns the distributions that generate it. The account is a reconstruction offered in the belief that the phenomenon is real and its classical foundation unsound, and its author remained a critic of capitalism on the strength of the relocated foundation. The second misreading holds that the work is a refutation of Marx. Its author presented it as a rescue: the ethical core of the critique, separated from a value theory he judged indefensible, and rebuilt on foundations economists and philosophers could inspect. Whether the rescue preserved the patient is the article’s question, but the intention was surgical rather than executionary, and assessments that ignore the intention misdescribe the episode’s place in the tradition’s history.

The third misreading holds that the generalised theorem shows exploitation is impossible, or that profit involves no transfer from anyone. The theorem shows that a particular argument for labour’s unique status fails. It leaves every other ground for exploitation claims standing, including the distributional ground its author built and the relational ground its critics defend, and it says nothing against the existence of transfers, which its own models compute in detail. The fourth misreading treats the withdrawal criterion as a policy proposal, as if Roemer had recommended that groups actually secede with their share of the capital stock. The criterion is a diagnostic instrument. Its policy descendant is coupon socialism, which achieves the counterfactual’s equal division inside the economy rather than by anyone’s departure, and confusing the instrument with the descendant inverts the relationship between the theory’s parts. The fifth misreading, the most consequential for students, holds that the labour theory of value is required for the ethical argument against capitalism, so that abandoning the theory means abandoning the argument. The whole construction exists to show the opposite: the ethical argument, re-founded on the injustice of asset distributions, proceeds without the value theory at every step, and a student who can state why the re-founding works, and what it costs, commands the material in a way that a student who merely repeats the classical derivation does not.

Did Roemer refute Marx or reconstruct him?

Reconstruct, by his own presentation and by the work’s structure. The books preserve the tradition’s conclusions about class and about capitalism’s injustice while replacing the foundations under them, and the replacement is announced as a service. The charge that the reconstruction changed the doctrine’s character is the critics’ point, and it is assessed on its merits rather than settled by the author’s intention.

The verdict: grading the relocation

The article’s governing claim can now be graded, and the grading follows the comparison basis fixed at the outset: what each account locates exploitation in, and what each can therefore explain. On location, the relocation is exactly as advertised. The withdrawal criterion derives every verdict from the asset distribution, requires no premise about production, and delivers class and exploitation as coincident classifications under the correspondence. On explanation, the account explains the incidence of class positions, the persistence of transfers across contractual forms, the sequence of grievances across social forms, and the reason profit and surplus labour correspond without labour being metaphysically privileged. These are substantial explanations, and the article’s assessment records them as successes without discount. On implication, the relocation moves the ethical argument to distributive justice, with the gains in tractability and the losses in relational content that the matched rows of the comparison table display.

The failure is equally specific, and it is a failure of capture rather than of logic. The account cannot capture domination: one person’s direction of another’s activity, inside production, as a constituent of the wrong. It cannot capture it because it contains no premise about production, and it contains no premise about production because its generality and tractability are purchased by that omission. The verdict of this article is therefore double, and the doubleness is the finding rather than a hedge. Judged by its handling of exploitation’s distributional roots, the relocation succeeds, and its success is permanent: no subsequent theory can responsibly ignore the demonstration that endowments generate the class structure, or the theorem that value accounting cannot privilege labour. As an account of exploitation whole and entire, the relocation fails, because the phenomenon the tradition named includes a relation the relocated concept defines out of existence. The failure is stated with its defeater, as this series requires of every verdict in the cluster. That failure half of the verdict falls the moment someone constructs the missing formulation, one whose test for exploitation looks inside production for a relation of direction and appropriation instead of looking only at endowments. Roemer’s framework does not contain such a formulation, the construction of one is invited by this article in the open, and until it is produced the failure half of the verdict stands.

How should the new account of exploitation be graded?

As a success on its chosen ground and a failure off it. The account masters the distributional basis of exploitation, proves the class correspondence, and dissolves labour’s formal privilege, while leaving domination inside production uncaptured, so the grade is divided exactly where the comparison table’s final row said it would be.

One closing observation concerns the pattern this article instantiates, which the series has met before. A concept is redefined onto a different basis, as the economic surplus was redefined in the monopoly capital literature, and the redefinition gains tractability while losing the content that made the concept distinctive. Naming the trade is the analysis, in that literature and in this one. Roemer’s trade was made knowingly, defended ably, and priced by its author more honestly than by most of its summarists. The reader who can state the trade in both directions, what the new basis buys and what the old basis kept, has the whole of the matter, and has it in the only form in which contested concepts can honestly be held: as a judgement with its conditions of defeat attached.

Property relations and the philosophy of holdings

The withdrawal account travels in the philosophical literature under a second name, the property relations theory of exploitation, and the name is worth adopting for a section because it states the account’s commitment without the machinery. Exploitation, on this theory, is a consequence of property relations: of who holds what, under what rules of transfer and inheritance, before any production begins. The name also locates the theory in its philosophical neighbourhood. The theory of justice in holdings, given its canonical modern form by Robert Nozick, asks whether a distribution is just by asking how it arose: just acquisition, just transfer, and rectification of past injustice exhaust the questions. Roemer’s criterion approaches the same territory from the opposite shore. Where Nozick’s test is historical, asking about the path by which holdings were assembled, Roemer’s test is patterned and counterfactual, asking about the distance between the actual distribution and an equal one. The two tests can disagree in both directions. A distribution assembled by impeccably voluntary transfers from an equal start can drift far from equality, and Roemer’s criterion will find exploitation in its later stages while Nozick’s finds none. A distribution produced by ancient conquest may, by demographic accident, sit close to equality, and the verdicts reverse. Neither test refutes the other. They encode different theories of what makes holdings just, historical entitlement against patterned equality, and the exploitation verdict inherits the encoding. Students who see this clearly stop asking which test is correct in the abstract and start asking which theory of justice they are prepared to defend, which is the question both tests were built to force.

The property relations name also clarifies the account’s relationship to a distinction philosophers of exploitation draw between distributive and relational wrongs. A distributive wrong is a wrong in the pattern of holdings itself. A relational wrong is a wrong in how parties treat each other within their interactions. Roemer’s theory is the limiting case of a distributive account: the relational facts drop out entirely, and the whole moral assessment is carried by the pattern. The domination theory is the limiting case at the other pole: the pattern matters only as it enables or expresses the relational wrong. Most working theories of exploitation in the wider literature are mixtures, holding that an unfair advantage taken within a transaction is exploitative when the transaction’s background is unjust and the advantage flows through the interaction. Roemer’s purity is what makes his theory scientifically interesting, since limiting cases test concepts the way extremes test materials, and it is also what makes the theory feel, to many readers, like a theory of something adjacent to exploitation, distributive injustice itself, wearing exploitation’s name. That feeling, articulated, is the deepest form of the emptied-concept objection, deeper than the supervision example, because it questions not a verdict but the theory’s subject matter. Roemer’s defenders answer that the subject matter was always distributive at root, the tradition’s relational language being the phenomenology of a distributive fact. The argument returns, as every path in this article returns, to the master question, and the reader’s answer to it will organise everything else the reader thinks about the account.

Why is the account called a property relations theory?

Because it makes exploitation a consequence of the relations of ownership in place before production begins. Who holds the productive assets, and on what terms those holdings can be transferred and inherited, determines who will exploit and who will be exploited, so the theory’s explanatory variables are all property variables.

There is a historical irony in the name that seminar rooms enjoy. Marx’s own vocabulary centres on relations, the relations of production, and the tradition that followed him used relational language as a badge of difference from bourgeois economics, which was accused of treating distribution as a natural fact about things rather than a social fact about people. Roemer’s relocation, executed by a Marxist with Marxist intentions, arrives at a theory in which the operative facts are holdings of things, assets in a column, and the relations between people are derived quantities. The irony is superficial in one respect: the asset distribution is itself a social relation, frozen at a moment, since an asset is only an asset within a system of recognised rights. Roemer would insist on exactly that. His models presuppose a regime of property rights as their institutional frame, and the distribution he measures is a distribution of rights over things, which is a relation between people mediated by things, as the classical tradition always said property was. The relocation thus does not abandon the relational ontology. It relocates the relation that matters, from the ongoing relation inside production to the standing relation constituted by the rights structure, and argues that the standing relation explains the ongoing one. Stated in that form, the dispute with the domination theorists becomes precise enough to be fruitful: both sides agree the wrong is relational; they disagree about which relation is fundamental, the constitution or the conduct it licenses.

The constitutional framing also explains the criterion’s otherwise puzzling silence about desert and contribution. A theory of holdings can ask whether the distribution arose justly, and egalitarians and libertarians answer differently, but neither needs to ask whether particular holders deserve their holdings by their efforts or talents, since holdings on both views are justified or not by the rules of their acquisition and transfer rather than by the moral qualities of their holders. Roemer’s criterion inherits this silence. The diligent accumulator and the inheritor stand identically in its columns if their asset positions match, and the leisure-preferring worker’s verdict does not consult her choices except as they shaped her endowment. Critics from the responsibility-sensitive wing of egalitarianism have pressed Roemer on exactly this point, and his later equality of opportunity work, with its circumstance and effort distinction, can be read as the framework growing the missing organ: an account of which inequalities trace to choice and which to circumstance, grafted onto the distributional rootstock the exploitation theory had grown. The graft matters for the assessment of the whole career. The relocation was not a resting place. It was the first member of a sequence of constructions, each repairing a silence its predecessor had been criticised for, and the sequence’s direction, from holdings to circumstances, is the direction of post-Rawlsian egalitarianism itself.

Needs, desert, and the claims the criterion does not hear

A benchmark built on equal endowments answers one question about justice and is silent about two others that any complete distributive theory must face: what people need, and what people deserve. The silence is worth a section, because readers sometimes fault the criterion for verdicts that belong to the other two questions, and the faulting confuses the instrument’s dials. Consider need. A coalition whose members include many dependents, the young, the old, the ill, may require a larger share of resources to reach the same welfare as a coalition of healthy adults, and an equal parcelling of assets per head will leave the first coalition worse off in welfare though identical in endowment. The withdrawal test, applied to assets, may therefore pronounce a needy coalition unexploited while its members live harder lives than an exploiting coalition of the healthy. Roemer’s framework absorbs the point by assigning need to a different department of justice: the criterion measures exploitation, one specific wrong, and a society’s obligations to need run on a separate ledger, through provision and insurance, whose design his later work on opportunity and responsibility addresses. The absorption is legitimate and the boundary it draws is real. Exploitation, on the relocated concept, is not a synonym for every distributive grievance, and a theory that made it one would lose the precision the relocation bought.

Desert raises the mirror difficulty. A coalition of the exceptionally industrious, whose members built their endowments by effort under fair starting conditions, will be pronounced exploiting if its members would lose from an equalising departure, and the verdict pays no attention to the effort’s moral weight. The framework’s answer parallels the answer on need: desert, like need, is a separate dial, and the egalitarian benchmark deliberately brackets it, on the ground that holdings in actual economies trace so little to unaided effort, and so much to inheritance and circumstance, that a test which waited on desert’s measurement would never return a verdict at all. Critics from the responsibility-sensitive tradition have pressed Roemer to build the dial into the instrument, and his later work on equality of opportunity, separating circumstance from effort, is the long answer. The exploitation criterion itself remains a one-dial instrument, and the article’s assessment has been framed accordingly: it is a theory of exploitation’s distributional basis, not a theory of distributive justice entire, and judging it as the second when it claims to be the first is the mirror image of the error its partisans commit when they wield it as a complete moral audit.

Why does the test ignore need and desert?

Because it was built to isolate a single variable, the endowment distribution, and to return verdicts about that variable alone. Need and desert are genuine claims on a distribution, but they are different claims, assigned by the framework to other departments of justice, and an instrument that tried to register all three at once would register none of them precisely.

The two silences complete the inventory this article has kept. The criterion is silent about production’s interior, about how holdings came to be assembled, about need, and about desert. Four silences is a great deal of quiet for one instrument, and a reader is entitled to ask whether anything is left inside the concept once the quiet is subtracted. What is left is the thing itself, on the relocation’s own terms: the demonstrable, computable, cross-form connection between who starts with what and who ends up commanding whose labour. The tradition before Roemer asserted that connection in the language of essence and appearance. Roemer proved a version of it in the language of equilibrium, and the proof survives every silence around it, because the silences were never denials. They were the boundaries of a claim, drawn by an author who preferred a narrow truth to a broad impression, and the preference, with all its costs itemised in this article, is the relocation’s final lesson in theoretical character.

The relocation pattern: the surplus precedent

The brief for this article notes that the relocation instantiates a pattern the series has met before, in the monopoly capital literature’s treatment of the economic surplus, and the precedent deserves a fuller comparison than a passing mention, because the parallel is exact enough to be diagnostic. In the monopoly capital construction, the surplus was redefined away from its classical basis in value magnitudes and onto a basis in observable accounting quantities, the gap between output and the costs a society must incur, in order to make the concept measurable in national statistics. The gain was tractability: the surplus could be estimated, its absorption traced through military spending, waste, and the sales effort. The loss was the classical concept’s connection to production relations, since an accounting gap, like an asset distribution, is silent about the relations inside the workplaces whose costs and outputs compose it. Critics of that construction made, in their day, the same charge this article has weighed against Roemer: that the concept had been made measurable by being made into something else.

The parallel runs deeper than the shared trade. Both relocations were performed by insiders as acts of rescue, against the background of a foundation judged unusable: value magnitudes that could not be measured in the one case, could not privilege labour in the other. Both produced instruments that travelled, the surplus into development economics and the critique of waste, the withdrawal test into distributive ethics and class analysis. Both provoked a defence of the older concept’s relational content from critics who valued the content above the tractability. And both left the field with a two-register vocabulary, in which careful writers specify which basis a given claim uses and careless ones slide between registers mid-argument. The pattern’s recurrence across unrelated corners of the tradition suggests it is not an accident of two authors’ temperaments but a structural temptation of the material itself. Marxian concepts were forged as fusions, part measurement and part indictment, part mechanism and part moral charge. Every generation of the tradition’s economists inherits the fusions and feels the pull to separate the parts, since separated parts can be proved, estimated, and defended in the republic of letters. Every separation is then mourned by the tradition’s other wing as an amputation, since the fusion was the concept’s rhetorical and political power. Roemer’s exploitation is the pattern’s purest instance because its author stated the trade most explicitly and paid the price most knowingly, which is why this article has treated the episode as the series’ model case for the reading of reconstructions generally.

A closing caution follows from the pattern’s recurrence, addressed to writers rather than to theorists. Sliding between registers mid-argument, invoking the measurable version of a concept when evidence is demanded and the fused version when indignation is wanted, is the characteristic vice the pattern makes available, and it is a vice of exposition before it is a vice of theory. The discipline this article has practised, naming the basis in play at each step and pricing each trade as it is made, is the only known prophylactic. Readers who apply the discipline to the next reconstruction they meet, in whatever field, will find the trades easier to spot for having watched this one itemised.

Skill, talent, and the exploitation that survives socialism

The most quietly radical element of the 1982 construction is not the capitalist definition but the socialist one, and it rewards a section of its own. Suppose the capitalist grievance were fully remedied: alienable assets held in common or distributed equally, so that no coalition could gain from the capitalist withdrawal test. Roemer’s taxonomy insists that a further complaint remains available, because people differ in inalienable endowments, in strength, dexterity, intelligence, training, and the marketable scarcity of their combinations. In an economy of equal alienable assets, the talented earn premiums, command better terms, and accumulate advantages over the untalented, and the untalented can pose the withdrawal question in its socialist form: would we be better off if inalienable endowments, too, were dealt out per capita? The counterfactual cannot be enacted, since talents cannot be detached and redistributed, but it can be posed, and posing it defines a standard against which even a post-capitalist economy can be found exploitative. Roemer drew the moral without flinching. Exploitation, in the generalised sense his taxonomy builds, is not a specifically capitalist pathology. It is the name for advantage extracted from any unequal endowment, and each social form is characterised by which endowments it permits to generate advantage and which it neutralises.

This result reframes the socialist project in a way that discomforted the tradition and that Roemer plainly intended. If socialism is defined by the abolition of exploitation, and exploitation survives the socialisation of alienable assets in its skill form, then socialism as an end-state recedes: the project becomes a sequence of neutralisations, each leaving a residue for the next stage to address, with no guarantee that the sequence terminates. The taxonomy’s fourth cell, status, extends the sequence into the territory of office and privilege, suggesting that a society which neutralised both asset and skill advantage could still generate exploitation through bureaucratic rank. Read as a whole, the taxonomy is a theory of why emancipatory projects keep discovering new grievances behind the ones they abolished, and it achieves this without any psychology of ingratitude or any cynicism about human nature. The grievances are structural. Each is generated by a real asymmetry in endowments that the reigning form leaves operative, and each becomes visible, as a grievance, only when the previous asymmetry has been neutralised and attention can turn. Roemer’s framework thus explains a pattern in the history of egalitarian movements that the classical account, with its single mechanism, handles less naturally: the pattern of successive radicalisations, in which the previous stage’s remedy becomes the succeeding stage’s grievance.

Why does the skill form of the test matter to the theory?

Because it proves the instrument is not a device for indicting capitalism alone. Applied beyond its home case, it finds exploitation in an economy of perfectly equal alienable assets, generated by unequal talents, so the theory’s real subject is revealed as unequal endowment as such, in whatever asset a social form leaves operative.

The skill form also exposes the criterion’s dependence on counterfactuals of increasing strain. The capitalist test imagines a redistribution that is physically possible, however politically remote. The skill test imagines a redistribution no one could physically carry out, and its verdicts therefore have a different epistemic character: they express a judgement about the moral arbitrariness of the natural lottery rather than a comparison between feasible arrangements. Roemer accepted the strain, arguing that the moral point, that nobody deserves their endowment of talents, survives the impossibility of the redistribution that dramatises it. Critics answered that a test whose central counterfactual cannot obtain has stopped measuring and started sermonising, and that the distance between the capitalist and skill forms of the criterion is the distance between economics and moral philosophy wearing economics’ clothes. The exchange matters for the assessment of the whole construction because it locates the exact point where the account’s scientific virtues, computability and checkability, begin to drain away: they belong to the alienable case, and the taxonomy’s further cells inherit the criterion’s form while progressively forfeiting its discipline. A defender can reply that the taxonomy is honest about the gradient, presenting the cells as a sequence of decreasing operationality. A critic can reply that the sequence’s later cells borrow authority from the first cell’s rigour that they have not earned. Both replies are available in the literature, and the article records the frontier rather than policing it.

There is a biographical sequel worth one paragraph. The problem of talent did not leave Roemer’s work when the exploitation theory was complete. It migrated into the equality of opportunity framework as the problem of circumstance: the talented adult’s premium traces partly to endowments she did not choose and partly to efforts she did, and a just society, on the view Roemer developed, compensates for the first while respecting the second. The exploitation theory had posed the problem in its starkest form, the talented as exploiters under the socialist test, a verdict Roemer never withdrew and never quite celebrated. The opportunity framework is in part an attempt to refine that verdict into something a policy could use, distinguishing the premium that rewards effort from the premium that rewards luck in the genetic and social lotteries. Readers who find the 1982 verdict about talent perverse and the later framework sensible are watching a theorist negotiate with his own theorem across twenty years, which is among the more instructive spectacles modern political economy offers.

Roemer and the labour theory of value: an insider’s farewell

The abandonment of the labour theory of value by one of the tradition’s leading economists requires careful statement, because farewells from inside differ from attacks from outside, and Roemer’s was unmistakably an insider’s. He had worked within the value-theoretic framework, contributed to its formal development, and proved results inside it before concluding that the framework was unnecessary for the purposes that mattered. The farewell’s logic has three steps, each of which this article has already assembled and which are gathered here for the student’s convenience. First, the fundamental Marxian theorem establishes a correspondence linking profit to surplus labour, and the correspondence is genuine. Second, the generalised theorem establishes that the correspondence is not unique to labour, so the correspondence cannot certify labour as the source of the surplus in any morally loaded sense. Third, the withdrawal criterion supplies exploitation verdicts without the value apparatus, so the ethical argument loses nothing when the apparatus is retired. The conclusion is not that the labour theory of value is false in every version and for every purpose, a claim about price theory that belongs to the theory’s own article in this series, the labour theory of value explained. The conclusion is that the theory is dispensable for the critique of capitalism, and that a critique which can dispense with a contested foundation should.

The reaction inside the tradition split along a line this article has met repeatedly. Economists formed in the Sraffian and post-Keynesian currents had largely reached parallel conclusions by other routes, and received Roemer’s farewell as confirmation. Philosophers and theorists for whom the value categories carried the tradition’s distinctive account of capitalist social form, the analysis of the commodity, of abstract labour, of the wage as a form that conceals its own content, received the farewell as the moment the analytical school disclosed its true allegiance: to the questions of distributive ethics, which any liberal could ask, over the questions of social form, which were Marxism’s alone. That charge deserves its weight. The relocation does render the theory of exploitation continuous with general egalitarian theory, and continuity has a cost in distinctiveness. A tradition that asks only the questions everyone asks will eventually be answered by everyone’s theories, and the specific gravity of Marx’s analysis, its insistence that the wage form itself is a mystification and that production’s interior is a political arena, is precisely what the relocation brackets. Roemer’s response to the charge, across his writings, is disarmingly frank: if the distinctive claims cannot be sustained by arguments that meet public standards of rigour, their distinctiveness is not a defence but a symptom, and a tradition confident in its ethical core should be willing to re-found that core wherever sound arguments lead. The reader must decide whether the frankness is intellectual honesty or a failure to understand what made the tradition worth reconstructing, and the decision will likely track the reader’s answer to the master question about domination.

On what grounds does Roemer set the labour theory of value aside?

Because his own theorems show it is unnecessary to the ethical argument. The generalised theorem proves the value accounting cannot single out labour, and the withdrawal criterion grounds exploitation claims in asset distribution without it, so retaining the theory adds a contested foundation under a conclusion that stands without one.

It is worth recording what the farewell did not include, since summaries often smuggle in extra items. Roemer did not conclude that Marx was unimportant, that the analysis of capitalism should cease, or that distributive questions exhaust social criticism. He continued to describe himself within the tradition’s broad church, served as an editor and interlocutor for its debates, and built his market socialism as an explicit contribution to the socialist tradition rather than as an exit from it. The farewell was addressed to a theory, not to a movement. Nor did the farewell depend on the empirical fortunes of the value theory as price theory. Even a reader who holds that labour values track prices tolerably well can accept Roemer’s argument in full, since the argument is that the ethical conclusion never needed the tracking. Conversely, a reader who rejects the value theory on price grounds has not thereby accepted Roemer’s relocation, since the domination theorists reject the value foundation and the distributional relocation together, rebuilding exploitation on relations instead. The logical geography has four positions, not two, and examinations reward students who can draw the map: value theory retained or abandoned, exploitation located in production or in distribution, independently variable, with Roemer occupying the abandoned-and-distribution cell and Vrousalis the abandoned-and-production cell, a pairing that surprises students who assumed the debates were one debate.

The criterion at work on history: feudalism and the shape of a research programme

A definition’s worth is tested by the work it organises, and the withdrawal criterion organised a distinctive body of historical and theoretical work in the decade after 1982, much of it concerned with transitions between social forms. The feudal application shows the instrument’s flexibility at its best. The classical debate about the transition from feudalism to capitalism, the Dobb and Sweezy exchange and its successors, turned on whether the motor of change was internal to feudal production relations or external, in trade and towns. Roemer’s feudal exploitation definition reframed the question in his idiom: feudal exploitation consists in the obligations extracted from direct producers who possess their own means of production, and a feudal economy’s dynamics can be studied by asking when the producers’ withdrawal option, departure with their own holdings freed of dues, becomes attractive enough to discipline or dissolve the lords’ extraction. The reframing does not settle the historiography, and this article, following series discipline, treats historical verdicts as belonging to the historical literature. What the reframing demonstrates is the criterion’s portability: a definition stated abstractly, in terms of groups, endowments, and counterfactual departures, can be carried into a historiographical debate and can reorganise its questions without presupposing its answers.

The transition work also generated the construction known as the feudalism to capitalism thought experiment in Roemer’s models, in which an economy of self-sufficient producers with their own assets is shown capable of differentiating into capitalists and workers through nothing more than differential accumulation and the opening of labour and credit markets. The construction matters beyond its historical suggestion. It demonstrates that the class structure requires no original act of expropriation to arise, no violent separation of producers from the land, if differential saving and market access are given time to work. The classical tradition’s narrative of primitive accumulation describes how the separation in fact occurred in the paradigmatic English case, and the description stands as history. Roemer’s construction addresses the modal question behind the narrative: whether separation by force was necessary for capitalism, or merely the historically chosen path among several. His answer, that it was not necessary, has the same double edge as everything else in the construction. It frees the critique of capitalism from dependence on a contested origin story, since exploitation in the present requires no crime in the past. It also dissolves the moral charge the origin story carried, the sense that the present distribution is the fruit of theft and stands condemned by its parentage. A patterned test cannot hear that charge. Whether the charge deserves a hearing is, once more, a question about which theory of justice in holdings the reader holds, and the article has already shown the two tests, historical and patterned, delivering their opposed verdicts on the same distributions.

How does the feudal case test the criterion’s portability?

By requiring the instrument to pronounce on an economy whose producers own their instruments and whose extraction runs through obligation rather than through a labour market. The feudal form of the test, departure with one’s own holdings freed of dues, returns sensible verdicts there, showing the definition travels across social forms rather than being built for one.

The research programme the criterion organised had a recognisable life cycle, and intellectual honesty requires the cycle to be completed in the telling. The first phase was construction and consolidation: the definitions, the correspondence, the theorem, and the extensions to credit, skills, and status. The second phase was application and engagement: historical transitions, class analysis, the exchanges with Cohen and Wright and the domination theorists. The third phase, visible from the nineteen nineties, was absorption and dispersal. The programme’s questions migrated into its founder’s work on political economy and equality of opportunity, into the class analysis literature’s pluralised asset maps, and into political philosophy’s standard menu of exploitation accounts, while the programme as a distinct enterprise, a Roemerian school of exploitation studies, never quite formed. Absorption of that kind is a common fate for successful formalisations. The results become part of the discipline’s furniture, cited without ceremony, and the founder’s name attaches to the theorems rather than to a movement. Readers assessing the relocation’s importance should count the absorption as evidence for it: ideas that are merely fashionable are discarded, while ideas that are absorbed have been found load-bearing by people with no loyalty to their origin.

Unfree labour and the criterion’s edge

Every definition of exploitation should be walked to the edge of the free labour market and made to pronounce on what lies beyond it, because the tradition’s concept was formed in a century when unfree labour, slavery, indenture, penal labour, debt bondage, was not capitalism’s prehistory but its contemporary. The withdrawal criterion’s behaviour at the edge is instructive. A slave owns nothing, not even, in law, her own labour power, and the economy she works in assigns every asset to others. Pose the test’s capitalist form for the coalition of the enslaved: departing with an equal share of the society’s alienable wealth, they would be immeasurably better off, and the criterion returns the exploited verdict at maximum strength. The verdict is correct and the route to it is revealing, since the criterion reaches slavery’s wrong through the asset column alone, registering the totality of the slaves’ expropriation as a distributional fact while saying nothing about the whip, the chain, or the sale of persons, which are the wrong’s substance in every account the enslaved themselves left. Roemer’s framework would classify those instruments under domination, a distinct wrong, and the classification is available and coherent. The edge case thus performs the article’s whole argument in one scene: the criterion’s verdict is right, its silence is total, and whether the silence invalidates the verdict is the master question wearing historical dress.

Indenture and debt bondage occupy the middle ground where the criterion earns its keep most distinctively. The indentured labourer has signed, has received consideration, an advance, a passage, a loan, and is bound for a term by the debt the consideration created. Formal contract analysis finds consent and consideration and stops. The classical account finds a labour relation and asks about surplus extraction within it. The withdrawal criterion asks the question that cuts deepest in such cases: what endowments did the parties bring to the signing, and would the labourer’s coalition gain from a world in which the signing had occurred among equals? Where the answer is yes, the criterion pronounces the arrangement exploitative however impeccable its paperwork, and it identifies the advance itself, the loan that created the bond, as the mechanism by which an unequal endowment distribution converts itself into a period of unfree labour. The credit-market isomorphism, which appeared earlier in this article as an abstract result about lenders and borrowers, returns here with historical weight: bondage through debt is the isomorphism lived, the lending relation and the labour relation fusing in a single institution, exactly as the model’s equivalence between the two markets would lead one to expect.

What does the criterion register about slavery, and what escapes it?

It registers the total expropriation, since a coalition owning nothing gains maximally from departure with an equal share, and the verdict is as strong as the test can return. What escapes it is the coercion itself, the violence and the sale of persons, which the framework assigns to domination as a separate wrong rather than to exploitation as it defines the term.

The edge cases also discipline a temptation in the criterion’s popular use, the temptation to treat its verdict as a complete moral audit of an institution. Slavery fails the audit’s every test in Roemer’s framework too, once domination is admitted as a wrong beside exploitation, and no defender of the framework has suggested otherwise. The audit metaphor itself is the problem. Frameworks in this field are instruments for seeing, and the demand that one instrument register every wrong in its field of view is a demand that no instrument in the social sciences meets, including the classical account, whose own silences, about the distribution that precedes production and about wrongs among the asset-holding classes, this article has noted in their place. The edge cases therefore close the exposition where the verdict section closed the argument: with a divided assessment, honestly itemised, and with the invitation standing to any reader who can build the instrument that registers both axes at once.

Status, office, and the fourth cell

The taxonomy’s fourth cell, status exploitation, is the least developed in Roemer’s own writings and the most suggestive for readers whose interests run to the bureaucratic societies of the twentieth century or to the internal hierarchies of large organisations in any century. The idea is spare. Where command over resources attaches to office rather than to ownership, the holders of high office enjoy a form of advantage that is neither asset-based nor skill-based: the official’s access to housing, transport, information, deference, and discretion flows from position in a hierarchy, and it vanishes when the position is surrendered, as an asset does not. A withdrawal test can be posed for this form as for the others. A group counts as status-exploited when levelling the perquisites and powers of office, or rotating them, would leave its members better off, the departure imagined being from the hierarchy’s privilege structure rather than from the economy’s asset structure. Roemer sketched the cell to complete the taxonomy’s logic, and the sketch’s brevity is itself informative: the cell is where the criterion’s operational discipline is weakest, since the equalisation of office privilege is harder to model than the equalisation of a divisible stock, and the counterfactual’s terms are correspondingly contestable.

The cell nonetheless does real work in the theory’s architecture. Without it, the taxonomy would imply that a society could abolish exploitation by equalising assets and somehow neutralising skill premiums, while leaving a nomenklatura’s privileges untouched and unnameable in the theory’s vocabulary. With it, the theory claims coverage of the full sequence of forms the tradition has criticised, and the claim of coverage is the generality on which the relocation’s prestige rests. Critics have treated the fourth cell as the taxonomy’s tell. A framework that must stretch its instrument to cover bureaucracy, they argue, and that can cover it only by thinning the instrument’s operational content, reveals the strain that the drive for generality imposes. The first cell’s rigour was purchased with a divisible stock and a computable counterfactual. The fourth cell has neither, and its inclusion suggests that the taxonomy’s unity is partly rhetorical, a family portrait in which the later children resemble the first mainly in the framing. Defenders answer that family resemblance, honestly labelled, is all the unity the taxonomy ever claimed, and that a map which marks its own decreasing precision is more trustworthy than one which pretends to uniform accuracy. The article records the exchange as it records the skill frontier: a genuine boundary of the construction, visible from inside it.

What distinguishes status exploitation from the asset forms?

Its basis in position rather than in holdings. The advantage flows from office, from rank in a hierarchy and the discretion and perquisites attached to rank, and it cannot be bequeathed, sold, or divided like a stock of assets, so the withdrawal test must imagine the equalisation of privilege rather than the redistribution of a divisible endowment.

For the student, the fourth cell has a practical use disproportionate to its development in the literature. It supplies the vocabulary for a class of examination and seminar questions that the first three cells cannot reach: questions about whether the societies that abolished private ownership of the means of production thereby abolished exploitation. The disciplined answer the taxonomy makes available is that abolishing one basis of exploitation leaves the others standing, that the asset test falls silent while the status test continues to register, and that a verdict on any historical case requires specifying which cell’s test is being applied and what the evidence shows under it. That answer is more valuable than any verdict it might be used to reach, because it converts a shouting match about a word into a sequence of checkable sub-questions. The conversion of shouting matches into checkable sub-questions is, in the end, the analytical school’s whole pedagogical offer, and Roemer’s exploitation theory is its most complete exhibit.

Applying the criterion: a discipline of questions

For readers who want to use the instrument rather than merely admire or indict it, the criterion imposes a discipline of questions that can be stated as a sequence, and stating the sequence is the most practical service this article can perform. Begin by naming the coalition. No exploitation claim under this framework is well formed until the group alleged to be exploited is specified, its membership rule stated, and the temptation to slide between groupings refused. Continue by specifying the asset base. Which assets are alienable in the economy under study, how are they distributed across the coalition and its complement, and what data would establish the distribution? Continue by constructing the counterfactual honestly. The departing coalition carries a per capita share of the alienable stock, organises production among its members with its own capacities, and is judged better or worse off by its members’ own valuations of consumption and leisure, not by the analyst’s. Continue by checking the verdict’s stability. Does it survive plausible variations in the coalition’s boundaries and in the counterfactual’s assumptions about trade and indivisible assets, or does it flip when the construction is nudged? A verdict that flips under small nudges is a report about the construction, not about the economy. Conclude by naming the matters the verdict leaves unexamined. Domination inside the coalition’s workplaces, insecurity, the texture of authority: the criterion is silent on all of it, and a competent application says so on its face.

Worked informally on a familiar case, the discipline shows its value. Consider the claim that the employees of a single celebrated firm are exploited by it. The criterion first asks which coalition is meant: the firm’s employees alone, the employees of the industry, the asset-poor of the whole economy? The verdicts may differ, since the firm’s employees might be comparatively well placed within the wider coalition, and the criterion forces the claimant to own the choice. It then asks what the chosen coalition would carry on departure, which requires an economy-wide asset distribution, not a tour of the firm’s premises. It then asks whether the coalition would gain, which requires modelling its members’ production among themselves, a question about their skills and the assets’ productivity rather than about the firm’s internal life. The firm’s celebrated workplace culture, its supervision practices, its celebrated or notorious management, never enter. A claimant who finds that the interesting facts have all been excluded has understood the criterion perfectly, and has also understood the objection to it. The discipline’s questions are the relocation, operationalised.

Why must an application name what the verdict does not cover?

Because the criterion’s silence about production’s interior is systematic, and silence misleads when it is not announced. An application that reports a coalition unexploited, without adding that direction and subjection inside its workplaces lie outside the test, invites readers to conclude that nothing wrongful occurs there, which the test never examined.

The discipline has a mirror use against the account’s own partisans. A defender who deploys the criterion to dismiss a domination-based complaint, on the ground that the complainants’ coalition would not gain from withdrawal, has committed the mirror error: the test’s silence about the complaint’s subject matter is treated as a finding about it. Roemer’s own practice in print is generally careful on this point, distinguishing the questions his instrument answers from the questions it does not, and the care is worth imitating because the instrument’s prestige tempts its users into imperialism on its behalf. Every powerful test has this temptation attached. The test defines a quantity precisely, the quantity becomes the concept in its users’ mouths, and the concept’s other senses are first neglected and then denied. The history of intelligence testing offers the standing warning from another field. The withdrawal criterion’s users, friendly and hostile alike, are best served by the discipline of the final question: name the verdict’s blind spots, every time.

Wages under the relocation: what a pay packet measures

Follow the relocation into the most concrete object in political economy, the wage, and its interpretation changes underfoot. In the classical reading, a wage is the price of labour power, tending toward the cost of reproducing the worker, and its relation to the value the worker creates is the scene of exploitation. In Roemer’s reading, a wage is an equilibrium price formed by the interaction of endowments and technology, and it measures the scarcity of the worker’s contribution given the distribution of everything else. Two workers performing identical tasks at identical intensity can receive different moral descriptions under the criterion if their asset positions differ, since the asset-rich worker’s wage is a chosen convenience and the asset-poor worker’s identical wage is the visible term of her exploitation. The pay packet is the same. The verdict differs, because the verdict was never about the packet. Critics find this consequence the relocation’s reductio in miniature: a theory of exploitation under which exploitation cannot be detected by examining the employment relation itself, however closely, has cut the concept loose from its evidence. Defenders find the consequence the relocation’s maturity: moral verdicts about arrangements were never going to be read off single transactions, and a theory that forces the analyst to examine the background distribution before pronouncing is a theory that has understood where verdicts come from.

The wage’s interpretation also governs the reading of wage movements, and here the two accounts generate different commentaries on the same events. A general rise in wages, in the classical commentary, may leave exploitation untouched or even intensified, if productivity rises faster and the surplus share grows. In the relocated commentary, a wage rise financed from profits redistributes income without altering endowments, and therefore leaves the exploitation structure standing, while a wage rise converted by its recipients into assets, houses, savings, stakes, alters the structure itself in proportion as the conversion endures. The relocated commentary thus draws a line the classical one does not, between income gains that evaporate in consumption and income gains that consolidate into endowments, and it predicts that the second kind, not the first, is what gradually dissolves class positions. Housing policy, pension design, and the taxation of wealth become, under this commentary, exploitation policy in the strict theoretical sense, a reclassification that the policy grammar section of this article has already drawn out and that the wage discussion grounds in the pay packet where citizens actually encounter the theory’s subject matter.

Why can identical wages carry different verdicts under the criterion?

Because the verdict attaches to the worker’s position relative to the endowment distribution, not to the terms of the individual bargain. A wage accepted by a worker with substantial reserves is one option among several she might have chosen, while the identical wage, accepted by a worker with nothing, is the price of her lack of alternatives.

There is a historical application of the wage discussion that the criterion’s partisans cite and its critics should answer. The long compression of wage differentials and the broadening of home and share ownership in the postwar decades, in several economies, coincided with a widely registered softening of class antagonism that purely income-based accounts struggle to explain, since income inequality by several measures moved less dramatically than the antagonism did. The relocated account has a ready explanation: the endowment distribution broadened, through housing, pensions, and mass education’s effect on the value of inalienable assets, and the exploitation structure softened as the correspondence between asset poverty and labour-market position loosened. The explanation is offered here as an illustration of the framework’s reach, not as a settled historiographical verdict, and it is contestable on the evidence like every such application. Its form, however, is the form of the framework’s best work: a puzzling social fact, a distributive mechanism proposed, and a standard against which the proposal can be checked. Critics who would dismiss the framework wholesale owe this genre of its applications an answer, just as its partisans owe the domination literature an answer of the kind this article has required throughout.

The fissured workplace: when the employer disappears

A development in the organisation of work that postdates Roemer’s main constructions tests his framework and the classical one asymmetrically, and the asymmetry is instructive. Lead firms increasingly shed direct employment, contracting production to subsidiaries, franchisees, agencies, and platforms, so that the worker’s labour is directed by an organisation that is not, in law, her employer, while the organisation that profits from her labour directs it only through contracts and standards set at a distance. Labour lawyers call this the fissured workplace. For the classical account, fissuring complicates the scene of exploitation without dissolving it: surplus labour is still extracted somewhere in the chain, but identifying the extractor, the party who appropriates and directs, requires tracing contractual layers designed to obscure exactly that identification. For Roemer’s account, fissuring changes almost nothing, and the insouciance is diagnostic. The criterion never located exploitation in an employment relation, so the relation’s legal fragmentation cannot hide the phenomenon from it. The asset-poor still sort into the chain’s lower tiers, the asset-rich still collect its returns, and the withdrawal comparison returns its verdicts indifferent to how many contracts intervene.

The asymmetry should not be scored too quickly in the criterion’s favour. Fissuring is not merely a legal veil over an unchanged reality. It is a technique for redistributing risk and discipline: the lead firm keeps the power to set pace, price, and standards while shedding the obligations of employment, and the shedding is itself an exercise of the relational power the domination account describes. A framework that registers no change when fissuring occurs has missed the phenomenon’s newest form, in which direction is exercised precisely by arranging not to be the director of record. The two-axis map absorbs the case without strain, as it has absorbed every case in this article: the distributional axis explains who is exposed to fissured work and who collects its margins, the relational axis explains what fissuring does to the experience and the legal visibility of subjection, and a complete account of the phenomenon needs both readings taken together. The case’s value for the student is that it is not a constructed counterexample but an institutional development, arising from the economy’s own evolution, which sorts the two accounts’ strengths as cleanly as any thought experiment in the literature.

Why does fissuring leave the criterion’s verdicts untouched?

Because the verdicts were never computed from the employment relation. Reassigning the worker’s legal employer among subsidiaries, agencies, and platforms changes who signs her contract while leaving her endowment, her alternatives, and her coalition’s counterfactual exactly as they were, so a test built on endowments reads through the rearrangement without noticing it.

There is a converse development worth setting beside fissuring, since it tests the classical account’s flexibility in the opposite direction. Where workers acquire significant assets through pension funds and housing, the classical account must still locate their exploitation inside production, in the surplus extracted during their working hours, and it does so without embarrassment: the extraction is a fact about the process, whatever the worker owns. The criterion, by contrast, must soften its verdict as the endowments broaden, until a sufficiently endowed workforce would not gain from departure and exploitation lapses by definition, however the working day is organised. Critics of the criterion treat this consequence as its quiet absurdity, a world of well-propertied but closely driven workers pronounced unexploited. Defenders treat it as the theory’s empirical wager: that a workforce with real reserves would in fact transform its working conditions through the bargaining power reserves confer, so the driven character of the work would not survive the endowments that dissolve the verdict. The wager is testable in principle, and the evidence of high-asset workforces in several economies, whose members have used their security to refuse the worst terms and shorten the reach of command, gives it more support than its critics usually allow. The exchange is left open here, as the article has left the frontier questions open, because its resolution belongs to evidence the exposition cannot supply.

Pension funds and the ownership that is not felt

One institutional development gives the criterion a natural experiment it has rarely been granted in the literature, and the experiment deserves a section because it separates ownership’s legal form from its lived substance. Peter Drucker argued in the nineteen seventies that pension funds had quietly made employees the beneficial owners of a large share of the capital stock in the advanced economies, a transformation he called pension fund socialism. If the workers own the means of production through their retirement accounts, has exploitation in Roemer’s sense withered? The criterion gives the question a disciplined form. What matters is not the aggregate share held by funds but the distribution of beneficial claims across persons, the degree to which the claims are alienable or locked until retirement, and the counterfactual gain each coalition would register from an equal parcelling. Measured that way, the pension transformation registers as a real but modest broadening of endowments: claims are spread far more widely than direct share ownership ever was, yet they remain steeply graded by earnings, inaccessible during the years when reserves would confer bargaining power, and controlled by fund managers whose decisions the beneficiaries do not direct.

The case is a gift to both axes of this article’s map. On the distributional axis, pension wealth is exactly the kind of endowment change the criterion exists to detect, and its softening effect on class positions, where it has occurred, is the framework’s kind of evidence. On the relational axis, the pension holder’s position inside her own workplace is untouched by her retirement account’s contents: she is directed as before, by managers answerable to funds that may include her own savings without including her voice. Drucker’s phrase claimed a revolution. The two-axis reading returns a more precise verdict, a genuine redistribution of claims without a redistribution of command, and the precision is the framework controversy put to work: each axis catches what the other misses, and the institution is misdescribed by either reading alone. Roemer’s coupon design can be understood, in this light, as pension fund socialism with the defects corrected in advance: claims equal rather than graded, universal rather than employment-linked, and constitutionally barred from reconcentration, though equally silent, as its critics never tired of noting, about command inside the firm.

Objections from the other direction

The objections heard so far came from the account’s left, from theorists who found it not Marxist enough. A complete hearing includes the objections from its right, from theorists who find its egalitarian benchmark indefensible, because the two sets of objections triangulate the account’s actual position in logical space. The libertarian objection is the fundamental one. The withdrawal criterion measures a coalition’s gain from an equal division of assets, and treats the gain as the signature of exploitation. But a gain from equal division exists whenever assets are unequal and the coalition is asset-poor, including when the inequality arose from free choices among equals, from gifts, from the differential success of enterprises that served consumers well. To call every such gain exploitation is to build the conclusion into the benchmark: equality of assets is assumed to be the just baseline, and departures from it are then discovered to be wrongful, in a procedure that would not persuade anyone who did not accept the baseline at the start. Roemer’s response, in his distributive writings, accepts the structure of the charge and denies its force as a refutation. The criterion, he in effect concedes, is not a proof that egalitarian distribution is just. It is an instrument for a theory that holds egalitarian distribution to be just on independent grounds, and within that theory it performs the indispensable service of connecting the abstract commitment to equality with the concrete classifications of class analysis. The concession clarifies the account’s logical geography at the cost of its apparent neutrality, and students should register the cost plainly: the withdrawal criterion cannot be used to argue an opponent into egalitarianism, only to show an egalitarian what follows.

A second objection from the same direction concerns the treatment of risk and time. Asset positions at any moment are snapshots of processes that run through time: saving, enterprise, insurance, and loss. Two coalitions with identical snapshots may have arrived by opposite routes, one by insuring prudently against risks that did not materialise, the other by gambling and being rescued by luck. The criterion, evaluating the snapshot against the equal-division counterfactual at the same moment, returns the same verdict on both, and a theory of justice sensitive to responsibility and luck will find the verdict coarse. This objection connects with the responsibility critique already noted from within egalitarianism, and Roemer’s later work on equality of opportunity is again the framework’s own attempt to refine the coarseness, by separating the components of advantage that trace to circumstance from those that trace to choice. The exploitation theory itself contains no such separation, and its defenders must either accept the coarseness as the price of a usable instrument or concede that the criterion states a sufficient condition for a distributive complaint rather than a complete theory of exploitation. Many careful defenders take the second option, and the article’s verdict has been framed to accommodate it: the relocation succeeds on the ground it chose, the distributional basis of exploitation, a formulation that leaves room for the basis to be one component of a fuller account rather than the whole.

What does the libertarian objection identify as the criterion’s hidden premise?

The egalitarian benchmark itself. Measuring exploitation by the gain from an equal division presupposes that equal division is the standard of justice, so the criterion cannot establish exploitation to anyone who rejects that standard, and its verdicts are conditional on a theory of distributive justice it does not itself supply.

A third objection, from neither direction but from the philosophy of social science, questions the counterfactual’s determinacy. Whether a coalition would be better off after withdrawal depends on what the coalition would do with its assets, which depends on its members’ choices in a world that does not exist, at prices that would form in markets that never meet. Roemer’s models discipline the counterfactual by computing equilibria, and within a model the verdict is determinate. Outside a model, in application to real economies, the counterfactual’s terms multiply: the departing coalition’s internal organisation, its trade with the remainder, the treatment of public goods and of the state’s functions, each choice moving the verdict. The objection concludes that the criterion is a model-relative notion masquerading as an economy-relative one, precise where it is least applicable and impressionistic where it is most needed. The defence available is the standard defence of model-based concepts throughout economics: the concept’s precision inside models is what allows its application outside them to be argued about productively, assumption by assumption, rather than traded as intuitions. The defence is sound as far as it goes. It concedes, in effect, that applying the criterion is an exercise of judgement informed by models rather than a measurement performed by them, and readers who meet the criterion in applied work should hold its verdicts at exactly that epistemic distance.

The benchmark defended: why equality is the test’s silent partner

Every application of the withdrawal criterion leans on a benchmark that the criterion itself cannot supply, and Roemer’s distributive writings amount to a long defence of that silent partner. The benchmark is equality of the alienable endowment: the departing coalition carries the same share as every other coalition of its size, and the verdict measures the distance between the world as it is and a world parcelled out in identical portions. Why should identical portions be the standard? The defence Roemer and the egalitarian tradition offer has several layers, and they are worth separating because critics often demolish one layer and declare the edifice fallen. The first layer is arbitrariness. The existing distribution of productive wealth descends from inheritance, conquest, enclosure, luck, and the accumulated outcomes of markets working on earlier distributions that descend from the same mixture. No one supposes the result tracks desert or need or any other morally legible pattern. A benchmark need not be the distribution justice finally recommends in order to serve as the standard against which a given arrangement’s tilt is measured, and equality is the natural zero of the scale, the point from which tilt in either direction can be read.

The second layer is the burden of proof. Treating equality as the benchmark does not assert that every departure from equality is indefensible. It asserts that departures require defence, that the holder of a larger share owes an account of the share’s size in a way the holder of an equal share does not. Defences can be given: shares may differ because their holders created the wealth, bore risks others declined, or accepted the inequality as the price of a system that leaves the worst endowed better off than equality would. Each defence is an argument, and the benchmark’s function is to make the argument necessary and to specify its subject. The third layer is diagnostic modesty. The criterion measures one thing, a coalition’s distance from the equal-parcel world, and the defence of the benchmark claims no more for the measurement than that the distance is morally relevant information, a reading on one dial of a panel that also holds dials for domination, for need, for desert, and for the historical manner in which holdings were assembled. Critics who attack the criterion for failing to be the whole panel are answered by the modesty of the claim, and partisans who wield the single dial as a complete audit are refuted by the same modesty, a symmetry this article has maintained throughout.

What work does the equality benchmark do inside the test?

It supplies the zero of the scale. Departures from an equal parcelling of the endowment are what the withdrawal comparison measures, so the benchmark converts a raw fact, that endowments differ, into a signed quantity, the direction and size of a coalition’s disadvantage, while leaving the defence of particular departures to arguments the benchmark forces into the open.

There is a subtler defence available, which Roemer’s game-theoretic formation made natural to him and which deserves recording because it answers the charge of arbitrariness from an unexpected direction. In cooperative game theory, the equal division is not the only candidate benchmark, and the theory’s solution concepts, the core, the bargaining set, the value, each embody a different standard of what a coalition may claim. Roemer’s choice of the per capita standard was a choice among these, made because the per capita standard is the one that asks nothing about the coalition’s bargaining strength, its threats, or its strategic position. A benchmark built from bargaining strength would certify as just whatever the strong can extract, laundering power into entitlement. A benchmark built from contribution would require a theory of contribution’s measurement, which is the value problem returning through the back door. Of the family’s standards, the per capita one alone is independent of both power and productivity accounting, and its independence is its defence: the test measures a coalition’s position against the one standard that neither the strong nor the productive can claim was rigged in their favour, because it was rigged in no one’s favour at all. Whether independence of that kind is what a benchmark of justice needs is, once more, a question in distributive ethics rather than in the theory of exploitation, and the criterion’s honesty consists in making the dependence visible instead of hiding it inside a definition.

Two axes: toward a map on which both accounts can be drawn

Step back from the controversy and a map suggests itself on which both accounts appear as partial projections of a larger territory. Let one axis measure the distributional basis: how far a group’s position is explained by the endowment it starts from. Let the other axis measure the relational reality: how far a group’s working life is governed by direction, appropriation, and subjection inside production. The classical account reads the second axis and infers the first, treating the extraction inside production as the phenomenon and the asset concentration as its precondition. Roemer’s account reads the first axis and derives the second’s population, treating the endowment distribution as the phenomenon and the relations as its expression. The domination account stays on the second axis and denies the first any independent moral standing. A complete theory of exploitation, if the territory admits one, would have to say how the axes interact: how endowments generate relations, how relations reproduce endowments across time, and which of the two a given complaint is actually about. Neither the classical texts nor Roemer’s construction supplies that complete theory, and the literature since 1982 can be read as a series of attempts to supply parts of it, Wright pluralising the endowment axis, Vrousalis formalising the relational axis, Cohen rebuilding the normative foundations under both.

The map also explains the controversy’s otherwise puzzling durability. Disputes persist when the parties are measuring different axes while using one word, because each party’s counterexamples are genuine on its own axis and irrelevant on the other’s. The worker supervised inside the collectively owned firm is the decisive case on the relational axis and a null case on the distributional one. The lender and borrower are the mirror image. Controversies of this shape are not settled by more examples. They are settled, when they are settled at all, by agreeing on the work the concept must do, and the work differs by context: a theory of justice needs the distributional axis, a theory of workplace politics needs the relational one, a history of capitalism needs the interaction. Roemer’s permanent contribution may prove to be less his verdicts than his demonstration that the axes are separable, that a rigorous account can be built on one of them alone, and that the price of the separation can be stated exactly. The demonstration changed what it is possible to say in this field, including for those who say it in order to refuse his conclusions.

Why do the counterexamples on each side fail to settle the dispute?

Because each side’s decisive cases are constructed on the axis that side measures. The supervised worker in a collectively owned firm registers on the relational axis and not on the distributional one, while the lender and borrower register in reverse, so the examples pass each other without engaging unless the parties first agree which axis the concept must track.

There is a lesson in the episode for the reading of theoretical controversies generally, and the series’ readers, who meet such controversies in every cluster, may take it as a transferable tool. When a reconstruction provokes the charge of evacuation, ask three questions. What did the old concept’s work consist of, itemised? Which items does the reconstruction still perform? Which items has it declined, and has it declined them openly or by sleight? Roemer’s reconstruction performs the classificatory work, the explanatory work about incidence, and the indicting work against distributions, and it declines the phenomenological work, the description of subjection from inside, openly and with reasons. Sleight is absent from the episode, which is why it remains the model case for teaching what reconstruction is. Concepts in a living tradition are not museum pieces, and they are not putty either. They are tools with histories, and the honest alternatives are to repair them in the open, as Roemer did, or to defend their inherited form against the repairs, as his critics did, knowing in both cases exactly which functions are at stake.

Game theory, coalitions, and the definition’s formal ancestry

The withdrawal criterion is sometimes labelled game theoretic exploitation, and the label records the definition’s ancestry in cooperative game theory, the branch of mathematics that studies what coalitions can secure for themselves. In that vocabulary, a coalition’s worth is what it can guarantee its members by acting together, and a distribution of the social product is stable, in one famous solution concept, when no coalition can do better by breaking away and taking what its own resources command. Roemer’s definition borrows the breakaway comparison and moralises it: the coalition that could improve its lot by breaking away with an equalised endowment, rather than with merely its own resources, is the coalition the existing arrangement exploits. The ancestry explains several of the definition’s features that puzzle first readers. It explains the focus on coalitions rather than individuals, since cooperative game theory’s primitive actor is the coalition. It explains the counterfactual form, since a coalition’s worth is inherently a counterfactual quantity. It explains the definition’s silence about process, since game theory’s characteristic move is to abstract from the sequence of moves to the structure of possible coalitions and payoffs. Readers who know the ancestry can predict the definition’s behaviour in new cases, which is the mark of having understood a formal construction rather than memorised it.

The ancestry also imports game theory’s characteristic difficulties. Coalition worth depends on what the coalition is assumed able to do, and in economic applications the assumption does the work: the departing coalition is credited with the ability to reproduce the economy’s production possibilities among its own members, an ability real groups possess only unevenly. Coalition formation itself is unmodelled; the definition tests coalitions the analyst selects, and the selection, as the earlier discussion of coalition choice showed, can carry the verdict. And the solution concepts of cooperative game theory were built to study stability and bargaining power, not justice, so the moralisation of the breakaway comparison is Roemer’s own step, taken for his own reasons, and not a deliverance of the mathematics. None of this diminishes the construction. It locates the construction’s discretionary joints, the places where judgement enters a structure that presents itself as pure mechanism, and a reader who can find the joints can argue with the construction at the level where argument is effective.

Why is the account sometimes called game theoretic exploitation?

Because its test is built from cooperative game theory’s central comparison, what a coalition could secure by breaking away. Roemer moralises that comparison by specifying the endowment the breaking coalition carries as an equal per capita share, so the definition’s coalition focus and counterfactual form descend from the mathematics it borrows.

A related formal ancestor deserves mention because it shaped the correspondence principle’s proof strategy. The core of an economy, in general equilibrium theory, is the set of allocations no coalition can improve upon by recontracting among themselves using only their own resources. Roemer’s models live in this neighbourhood: his agents sort themselves until no one can gain by changing strategy, and the class structure is read off the resting point. The neighbourhood matters for one substantive reason. Core theory contains famous results about large economies, in which the resting points of coalition-proof allocation coincide with competitive equilibria, and those results reassured Roemer’s first readers that his constructions were not exotic artefacts but applications of the discipline’s central machinery to the tradition’s central questions. The reassurance was part of the relocation’s rhetorical strategy. Nothing in the argument required the reader to learn a new mathematics, only to watch the old mathematics pointed at a new target, and the pointing, Roemer wagered, would make the results undeniable by anyone who accepted the machinery for other purposes. The wager largely succeeded within economics. Its failure to persuade the tradition’s philosophers is the other half of the story this article has told, and the halves together compose the relocation’s historical significance: a proof that a concept can be made rigorous and portable at the price of its home tradition’s recognition.

Exit, voice, and the politics of the withdrawal metaphor

The withdrawal at the criterion’s heart has a famous cousin in social science, and the kinship clarifies both. Albert Hirschman’s Exit, Voice, and Loyalty, published in 1970, distinguished two responses to decline or injustice inside an organisation: exit, the act of leaving, and voice, the act of protesting from within. Hirschman’s subject was the strategic choice between the responses and the conditions under which each disciplines an organisation. Roemer’s withdrawal resembles exit with the politics removed and the accounting added: the departing coalition in his construction does not actually leave, does not intend to leave, and exerts no pressure by the threat of leaving. Its departure is a measuring device. Yet the kinship is not superficial, because both constructions rest on the same insight, that a party’s treatment inside an arrangement is connected to the quality of its alternatives outside the arrangement. Hirschman’s loyal member endures bad treatment because exit is costly and voice is hopeful. Roemer’s exploited worker endures the wage relation because the alternative the asset distribution offers her is worse, and the criterion measures exactly the gap Hirschman’s framework leaves qualitative: how much better the outside would have to be, under equalised endowments, before endurance stopped being rational.

The comparison also exposes a question the criterion’s design suppresses. In Hirschman’s world, the possibility of exit, when it becomes credible, changes behaviour inside the arrangement: organisations that fear departure moderate their treatment of the members they wish to keep. In actual labour markets, the same mechanism operates under the name of bargaining power, and it distributes its benefits unevenly in a way that maps closely onto Roemer’s variables. Workers with assets, savings, credentials in demand, a second household income, can credibly threaten departure and are treated accordingly. Workers without reserves cannot, and are not. The withdrawal criterion, read dynamically, is a theory of this unevenness: the counterfactual equalisation it imagines is precisely the condition under which every worker’s exit threat would carry the same weight, and the distance from that condition is the distance between the bargaining positions real workers occupy. Roemer did not develop the bargaining reading, since his models are equilibrium constructions rather than bargaining models, but the reading is available in his materials, and it gives the criterion a use in industrial relations that its author left unexploited.

How does Hirschman’s exit differ from Roemer’s withdrawal?

In function. Hirschman’s exit is a real option whose credibility disciplines an organisation from outside, a political mechanism exercised by members who might actually leave. Roemer’s withdrawal is an imagined departure used to measure a distribution, a counterfactual that no coalition performs and whose force is analytical rather than strategic.

Voice, the other Hirschman response, has no counterpart in Roemer’s construction at all, and the absence rounds out the account’s portrait. The framework’s agents do not protest, organise, strike, or deliberate. They optimise. Their collective action problem, the problem of how a coalition of the asset-poor might form and act as a coalition, is assumed away by the definition’s grammar, which tests coalitions as if they were already constituted. The classical tradition, whatever its other faults, placed the constitution of the collective actor at the centre of its politics: class in itself becoming class for itself was its name for the transformation the definition assumes. A theory of exploitation that cannot represent the transformation will always need a companion theory of politics, and Roemer’s later writings on democracy and political competition can be read as the companion under construction. The criterion tells a movement what is wrong and how wrong it is, measured in endowments. It does not tell the movement how it comes to exist, and movements that mistake the measurement for the politics will be surprised by the world’s refusal to rearrange itself in accordance with a theorem.

The theorem’s assumptions, and what survives their relaxation

Theorems are contracts: conclusions in exchange for premises, with the premises printed in the fine print. The generalised theorem’s fine print deserves a reading, because popular presentations omit it and because Roemer’s own assessment of his result depends on which premises are doing the work. The fundamental theorem and its generalisation are proved for economies with a definite structure: produced commodities are made from produced commodities and the singled-out input by techniques that can be represented in an input-output table, the economy can reproduce itself, and prices are uniform. Within that structure the results are exact. The structure excludes, or handles only awkwardly, several features of real economies: joint production, where one process yields several products inseparably; fixed capital that survives many production periods; land and other non-produced resources; and labour that is heterogeneous in skill, which the classical tradition handles by reduction coefficients and Roemer’s framework handles by moving skill into the endowment column. Each exclusion has generated a technical literature, and some of that literature qualifies the classical theory’s results in ways its defenders and critics still dispute. For the generalised theorem’s philosophical purpose, however, the qualifications matter less than they appear to. The theorem is a possibility proof about arguments: it shows that an argument of a certain form cannot privilege labour, by exhibiting the same form privileging steel. Relaxing the assumptions complicates both versions of the argument symmetrically, and no relaxation has been produced that breaks the symmetry in labour’s favour, restoring the privilege the theorem removed. The burden the theorem imposes is therefore stable: anyone who wishes to restore labour’s privilege must supply an argument of a different form, resting on labour’s human character rather than on the accounting, which is exactly the relocation of the debate Roemer intended to force.

What kind of result is the generalised theorem, logically?

A symmetry result. It shows that a family of arguments shares a form that is indifferent to which basic commodity occupies the singled-out position, so any conclusion drawn from the form alone attaches to every occupant. Privilege for one occupant must therefore come from outside the form, from premises about that occupant’s distinctive nature.

The correspondence principle’s fine print repays the same attention. Its proof assumes agents who optimise with full information, markets that clear, and endowments that are given and enforceable. Relaxing the assumptions moves the result in instructive directions. Imperfect credit markets, in which the asset-poor cannot borrow against future earnings, strengthen the correspondence’s grip, since the poor are confined to selling labour even when enterprise would pay them better. Monopoly and monopsony, in which single buyers or sellers face the market, disturb the result’s cleanliness by letting position depend on market structure as well as endowment, and Roemer’s framework must then share explanatory credit with the theory of market power. Information asymmetries, in which employers cannot observe effort and must design incentives, reintroduce the interior of the firm into the explanation of positions, since the contracts offered to the asset-poor are shaped by monitoring problems the endowment alone does not determine. The pattern across relaxations is the pattern this article has found everywhere: the distributional basis remains visible in every version, and in every version something else joins it. The mature reading of the correspondence is therefore not that endowments are everything but that endowments are the load-bearing member of a structure with other members, and that Roemer’s proof, by isolating the load in a stripped model, made it possible to see how much of the building it carries. Assessments that demand the stripped model be the whole building mistake the method of isolation for a claim of completeness, a mistake Roemer’s critics sometimes make and his popularisers make more often.

Freedom to lose: the worker’s choice reconsidered

Return to the figure who has stood at the margin of several sections: the worker who chooses her employer, prefers the bargain to her alternatives, and is pronounced exploited by a test that never consulted the inside of her working day. The figure concentrates the account’s provocation, and Roemer’s Free to Lose was written largely to face readers like her in imagination. The title’s argument can now be stated with the theory behind it. The worker is free in the sense the liberal tradition prizes: no one compels her to sign this contract rather than that one, and the law protects her refusal of any particular master. She is unfree in the sense the withdrawal test measures: her option set is so constricted by her endowment that every option in it involves selling her labour to someone, and the collective position of her class, as Cohen’s argument showed, cannot be escaped by the class as a whole. Roemer’s contribution to this old debate is to refuse the question’s usual form. Instead of asking whether the worker is free or unfree, as if freedom were a single property the case must possess or lack, the framework asks what her position would be under a just distribution, and treats the distance between the positions as the morally operative quantity. Freedom talk, on this approach, is a sometimes misleading vernacular for a distributive fact, and the vernacular can be retired wherever the fact can be stated directly.

The retirement has a cost that Cohen, again, was positioned to name. Some unfreedoms are not distances from a just distribution. A worker with ample assets who is forbidden by law from changing employers, bound to a firm by a legal tie, suffers an unfreedom that no asset redistribution would touch, and a framework that translates every unfreedom into distributive distance will mistranslate it. Roemer’s taxonomy can gesture at the case through its feudal cell, since legal bondage is the feudal form’s essence, and the gesture is not nothing. But the case’s moral character, a direct restriction on a person’s movement imposed by another’s command, is relational through and through, and the framework’s translation of it into a withdrawal comparison loses the character while preserving the verdict. The pattern is by now familiar enough to state as a law of the relocation: verdicts survive translation, characters do not. Whether verdicts without characters are enough depends, for the final time in this article, on what the concept is wanted for. A court needs verdicts. A movement needs characters, the felt texture of the wrong that makes people act. Roemer wrote for courts of theory. The tradition he reconstructed also needed to move people, and the tension between the two needs runs under every page of the debate his work provoked.

What does the freedom to lose phrase claim about the worker?

That her freedom is real and worthless in the same act. She may refuse any employer, and the refusal of all employers is destitution, so the choice among masters coexists with the necessity of having one, and the necessity, not the choice, is what the asset distribution explains and what a just distribution would remove.

The phrase has had a career beyond its book, travelling into discussions of precarious work and of the gig economy’s contractual forms, usually without its author’s framework attached. The travel is understandable, since the phrase names an experience that contractual freedom’s defenders and critics both recognise, the experience of choosing under a compulsion that no individual counterparty imposed. Roemer’s framework gives the experience an address: the compulsion’s source is the distribution of assets, and its measure is the withdrawal comparison. Detached from the framework, the phrase becomes a lament. Attached to it, the phrase is a diagnosis with a remedy implied, the equalisation of the starting position, and the difference between a lament and a diagnosis is the difference the whole relocation was built to make. Readers meeting the phrase in the wild can now restore its framework, and with it the argument the phrase compresses.

From coalitions to classes: the definition’s second form

A technical feature of the 1982 construction deserves exposition because it explains how a definition stated for arbitrary coalitions becomes a tool of class analysis. The withdrawal definition, in its general form, applies to any coalition whatever, and an economy contains an astronomical number of coalitions, most of them sociologically meaningless. Roemer showed that, under the model’s conditions, attention can be restricted to the coalitions the economy itself generates: the classes. The reason runs through the correspondence. If the exploited are exactly those who sell labour and the exploiters exactly those who hire, then the sociologically operative coalitions are the classes defined by labour-market behaviour, and testing other coalitions, however diverting, adds no verdicts that matter for the theory’s purpose. The definition thus has two forms. The general form is the philosopher’s instrument, precise and unwieldy. The class form is the sociologist’s instrument, the same test applied to the groupings the economy produces. Much confusion in the secondary literature comes from quoting the general form’s verdicts about exotic coalitions as if they were the theory’s claims about classes, or from demanding of the class form a precision that only the general form possesses. The counterintuitive cases, in particular, divide by form: some arise only for exotic coalitions and lapse when the test is restricted to classes, while others, the talented under the socialist test chief among them, arise in the theory’s core applications and must be owned, as Roemer owned them.

Why can the test be restricted to classes without loss?

Because the correspondence aligns the verdicts with labour-market behaviour. The coalitions whose withdrawal verdicts carry the theory’s claims are the ones the asset distribution itself sorts into existence, hirers and sellers of labour, so testing the classes captures the exploitation structure while exotic coalitions add curiosities rather than findings.

The two forms also mark the boundary between the theory’s two audiences and its two kinds of critic. Philosophical critics engage the general form, constructing coalitions and cases to probe the definition’s edges, and their probes are fair game because a definition that yields absurd verdicts anywhere has a defect somewhere. Sociological and economic critics engage the class form, asking whether real economies sort their populations as the correspondence describes, and their evidence is the class structure research that Wright and others conducted. Roemer’s construction is unusual in exposing both flanks deliberately, and the exposure is a consequence of its ambition: a general theory, as the founding title promises, must survive as philosophy and as social science simultaneously. Most theories in this field choose one tribunal. The choice of both explains the volume and the longevity of the critical literature the book attracted, since every critic could find a flank congenial to their weapons, and it explains the book’s survival as well, since a construction attacked from two sides for four decades, and still assigned, has demonstrated a robustness that single-tribunal theories rarely need to prove.

Why the steel example persuades, and where its persuasion stops

The generalised theorem is usually taught through the steel substitution, and the pedagogy is worth examining because it reveals both the argument’s strength and its boundary. The substitution persuades by forcing a choice. Confronted with the parallel construction, a defender of labour’s privilege must either accept that steel is exploited, which is absurd, or identify the premise that labour satisfies and steel does not, which is the beginning of wisdom. The premises candidates offer are instructive. Labour is inseparable from the labourer, so its transfer involves the person. Labour is not produced for profit in the way steel is, or not only so. Labour’s expenditure is the spending of a life, hour by hour, in a way that steel’s consumption is not. Each candidate premise is about the human character of labour, and each, once stated, can be assessed directly, without value theory. The theorem’s pedagogical function is thus to flush the real argument into the open. Whatever survives the flushing, and something does survive, since the human character of labour is a fact and not a confusion, becomes the actual foundation of the classical intuition, and Roemer’s wager is that the surviving foundation is distributive: what matters about labour’s humanity, for the exploitation verdict, is that labourers are persons whose holdings can be unjust. The domination theorist’s counter-wager names a relational foundation instead: what matters is that labourers are persons whose activity can be commanded. The steel example cannot decide between the wagers. It can only, and this is no small thing, prevent both sides from pretending the value accounting had already decided.

Seminars discover a further boundary when a resourceful student proposes a commodity that is not absurd in the steel way. Consider land, or consider a scarce natural resource whose owners extract rents from an economy. The parallel construction applied to such a resource yields verdicts that many readers do not find absurd at all: there is a long tradition, running through Henry George and into resource economics, of treating rent on scarce natural assets as a transfer crying out for socialisation, and the generalised theorem’s machinery, applied there, formalises that tradition’s intuition rather than ridiculing it. The episode teaches the theorem’s true scope. The construction ridicules the singling out of an input when the singling out rests on form alone. Where an independent moral argument exists for treating an input’s owners specially, as Georgeans argue for land, the construction becomes a tool of that argument rather than its refutation. Roemer’s framework absorbs the lesson without strain, since on his account every such argument is at bottom about the distribution of the asset in question, land included, and the withdrawal test applies to landholding coalitions as to any other. The steel example’s absurdity, it turns out, was never in the mathematics. It was in the absence of a moral patient. Steel has no owners with claims and no welfare of its own, and the moment the substituted commodity has either, the parallel argument starts doing honest work.

What must a defender of labour’s privilege supply after the theorem?

A premise about labour that no other commodity satisfies, stated in terms of labour’s human character. The theorem closes the formal route to privilege, so the defence must argue from what labourers are, persons whose activity is their life, and that argument then competes directly with the distributional and relational accounts on open ground.

The theorem’s boundary has one more marker, placed there by its author’s own candour. The generalised result is a result about a style of argument, the style that infers moral privilege from an accounting correspondence. It does not touch arguments for labour’s distinctiveness that never used that style, and the classical corpus contains such arguments in its analysis of the labour process, of cooperation and manufacture, of machinery and the working day. A reader who concludes from the theorem that Marx’s account of exploitation has been refuted wholesale has mistaken a theorem about arguments for a verdict on a library. Roemer, to his credit, rarely made that mistake in print, distinguishing the dispensability of the value foundation from the value of the tradition’s other analyses. His summarists have been less careful, and the correction belongs in any complete exposition: the theorem retires one foundation. The building the tradition erected has other foundations, and assessing them is other articles’ work.

The firm as a government: the relational axis returns

The relational axis that Roemer’s framework brackets has, in the decades since the relocation, acquired its own distinguished literature, and one work in that literature can stand here for the axis’s continuing vitality. Elizabeth Anderson’s Private Government, published in 2017, argues that the modern workplace is a form of government, a system of command in which most adults spend half their waking lives subject to orders they had no part in making, enforced by sanctions up to exile from the community of the firm. Anderson’s targets are the celebrants of contractual freedom, and her argument proceeds by taking the employment contract’s own terms seriously and asking what they authorise. Read beside Roemer, the book is instructive in both directions. It demonstrates that the relational axis supports a critique of capitalism’s characteristic institutions that never mentions asset distributions, a critique whose force a reader can feel without accepting any egalitarian benchmark. It also demonstrates, by its silences, what the relational critique cannot do alone: Anderson’s workers are governed badly, but the question of why these people are the governed and those the governors, the question of the sorting, is one her framework raises and Roemer’s answers. The two literatures are complements wearing the clothes of rivals, and the scholarly temptation to crown one of them is a temptation this article has resisted throughout, on the ground that the map with two axes describes the territory better than either axis describes it alone.

The complementarity has a practical corollary for readers who work on real institutions. A reform programme built only on the distributional axis, equalising assets while leaving workplace government untouched, would satisfy Roemer’s criterion and leave Anderson’s complaint standing in full. A programme built solely on the relational axis, democratising firms while leaving the asset distribution untouched, would satisfy the domination critique’s institutional demand while leaving the sorting mechanism, and the inequalities it regenerates, in operation. The historical reform programmes that have attracted serious support tend to mix the axes, pairing wider ownership with voice inside firms, and the mixture is usually defended piecemeal, as politics. The theoretical point the mixture illustrates is that the concept of exploitation, in its full historical extent, was always a two-axis concept, and the analytical achievement of the period this article covers was to separate the axes cleanly enough that their mixture could become a deliberate choice rather than a confusion. Roemer performed the separation from one side. His critics, by insisting on what the separation lost, performed it from the other. The concept emerged from the controversy clearer than it entered, which is what theoretical progress in a contested field looks like.

What does the private government argument add to the domination objection?

An institutional description. Where the objection spoke of direction and subjection abstractly, the workplace-as-government argument specifies the sanctions, the unaccountability, and the absence of voice that constitute subjection inside firms, giving the relational axis a detailed anatomy that a purely distributional account has no vocabulary to reproduce.

There is a final disciplinary observation to make about the two axes, concerning the kinds of evidence each invites. Distributional claims invite measurement: ownership surveys, wealth shares, mobility tables, the whole apparatus of inequality statistics, in which Roemer’s framework sits comfortably and to which it contributed. Relational claims invite description: ethnographies of workplaces, legal analysis of managerial prerogative, testimony about the experience of command, evidence that resists aggregation and whose authority is of a different kind. A field that must hold both axes will always have a methodological fault line running through it, between the measurable and the describable, and the controversies this article has chronicled are surface expressions of that fault line. Recognising the line does not reconcile the parties. It does predict where their arguments will fail to engage, and it equips the reader to translate between them, which is the most that honest exposition can promise.

What do bosses do? The labour process literature answers back

The relational axis has an empirical literature as well as a philosophical one, and Roemer’s construction appeared in the same decade that gave that literature its modern classics. Harry Braverman’s Labor and Monopoly Capital, published in 1974, argued that capitalist management systematically separates conception from execution, gathering knowledge of the production process into managerial hands and reducing the worker’s role to the performance of specified motions. Stephen Marglin’s essay asking what bosses do, published in the same period, pressed a sharper question: if hierarchy in production were merely a technical necessity, a requirement of efficient coordination, then the boss’s function would be explicable by engineering. Marglin argued that hierarchy’s historical function was instead to secure the capitalist’s claim on the surplus, by fragmenting tasks so that no worker commanded a saleable whole product and every worker needed the capitalist’s organisation to work at all. The two arguments share a premise that Roemer’s models deny by omission: that the interior organisation of production is itself a social technology, chosen for its effects on power and appropriation, not a neutral wrapper around a production function.

Set Marglin’s claim beside the correspondence principle and the two illuminate each other’s blind spots with unusual precision. Marglin explains why production is organised hierarchically when flatter organisations are technically feasible: hierarchy manufactures the dependence that makes the wage relation work. Roemer explains why, hierarchy or no hierarchy, the people at the bottom of the organisation chart are predictably the people who arrived with nothing: endowments sort them there. A Marglinite reading of Roemer says the models assume as a backdrop the very institutional arrangement whose existence is the scandal. A Roemerian reading of Marglin says the history of organisational forms, however scandalous, leaves the distributive question untouched, since a worker cooperative movement that seized the factories while leaving the wider ownership structure intact would reproduce insider and outsider classes within a generation. Both readings are correct about the other’s limits, and the conjunction of the two literatures is the strongest form of the two-axis map: organisational power explains the mechanism of extraction, distributive position explains its incidence, and neither explanation metabolises the other.

What does Marglin’s question expose about hierarchy?

That hierarchy requires a justification it rarely receives. If bosses merely coordinated technically necessary cooperation, their authority would be an engineering fact. Marglin argued the authority historically served appropriation, fragmenting work so that producers could not market a whole product independently, which makes the firm’s interior a designed relation rather than a natural one.

The labour process literature also supplies the domination objection with its most concrete evidence, the evidence of the working day as lived. Ethnographies of factories and offices document the pacing of work by machinery and by quota, the surveillance of breaks and of conversation, the grading of workers by metrics they cannot inspect, and the peculiar exhaustion of labour performed at another’s tempo. None of this evidence is legible to the withdrawal criterion, which would pronounce on the ethnographer’s subjects by consulting the ownership statistics of their economy. Roemer’s defenders have a reply that should be stated fairly: the criterion was never offered as a phenomenology, and condemning it for failing to describe the working day is like condemning a map for failing to describe the weather. The reply is sound and insufficient in the same breath, because the tradition’s concept was a phenomenology among its other functions, a way of naming what the working day feels like from inside and why the feeling is a political fact. A map of the territory that omits the weather is a fine map. A tradition that loses its weather reports loses its ability to say why the territory matters to the people who live in it, and the loss registers in the tradition’s rhetoric long before it registers in its theorems.

The household: unpaid labour and the criterion’s boundary

The domestic labour debate, which ran through Marxist feminism in the decades around Roemer’s construction, posed a question neither the classical account nor the relocated one was built to hear, and the question marks a boundary of both. Households run on labour that no wage purchases: cooking, cleaning, care of children and of the old, the maintenance of the workers themselves. The classical account, centred on the wage relation, struggled to classify this labour, since no capitalist appropriates its product directly and no surplus value is realised through its sale, yet the labour is indispensable to the reproduction of the workforce and is distributed by gender with a regularity no theory of class could ignore. Feminist critics argued that a theory of exploitation silent about the household had defined its subject by the male biography, the worker who leaves home for the factory, and had nothing to say about the labour that sent him there fed and rested.

Roemer’s framework hears the question differently, and its answer is partial in an instructive way. The criterion can be applied within the household’s economy only awkwardly, since the household’s assets are typically held jointly in law while being controlled unevenly in fact, and the withdrawal the test imagines, a partner departing with a per capita share of the household’s wealth, is a real legal event in divorce rather than a counterfactual, which changes the test’s character. Where the framework does speak is at the economy’s boundary: the partner who performs the unpaid labour typically holds fewer alienable assets in her own name, faces the labour market with interrupted experience and diminished credentials, and belongs to coalitions the criterion pronounces exploited for reasons that run through the household without ever naming it. The framework thus registers the consequences of the domestic division of labour in its asset columns while remaining silent about the division itself, its daily allocation of tasks, its negotiations and coercions. The pattern is the article’s pattern once more, at a new site: verdicts downstream, silence upstream, and a critic, here the feminist literature, insisting that the upstream arrangement is the phenomenon. The insistence deserves its place in the assessment, because it shows the domination objection is not one objection but a family, arising wherever work is directed and appropriated outside the asset register’s field of vision, in the factory, in the office, and in the home.

What can the criterion say about unpaid domestic labour?

Only what follows from its traces in the asset columns. The partner who performs it usually accumulates fewer holdings in her own name and reaches the labour market with weakened credentials, so the coalitions she belongs to return exploited verdicts, while the household’s internal allocation of tasks and authority, the phenomenon the feminist critics named, lies outside the test’s construction.

The household case adds one refinement to the two-axis map that the factory cases did not require. Inside a household, the distributional and relational axes are not merely two readings of one structure; they are causally entangled in both directions within a single institution, since the relational allocation of unpaid tasks produces the distributional inequality of the partners, and the distributional inequality then underwrites the relational allocation’s stability. A theory that reads only one axis will therefore not merely omit part of the story but will misread the part it keeps, treating as a starting endowment what is in fact a produced outcome. The point generalises beyond the household, and Roemer’s defenders should concede it cleanly: endowments are not nature. They are the sediment of previous relations, and a framework that takes them as given for the purpose of a test must not let the test’s users forget the sedimentation. The criterion’s own author, whose historical work on transitions showed endowments being made and unmade by accumulation and expropriation, never forgot it. The forgetting is a vice of the framework’s consumers, and the household is the case that exposes it most reliably.

Predistribution: the relocation’s policy grammar

A theory that places the wrong in the starting distribution generates a characteristic policy grammar, and the grammar has become familiar enough in public debate that its theoretical source is often forgotten. Policies divide, in this grammar, into those that correct outcomes after markets have run, transfers and taxes on income, and those that shape the endowments with which people enter markets in the first place: education, inheritance rules, access to capital, the structure of ownership itself. The second family acquired the label predistribution in policy discussion, and whatever the label’s provenance, the family’s theoretical charter is the relocation this article has examined. If exploitation is a function of the asset distribution, then the deepest anti-exploitation policy is the one that changes the distribution of assets, and income transfers, however generous, treat the flow while leaving the stock, and the class structure the stock generates, untouched. Roemer’s coupon socialism is the grammar’s most radical sentence, a constitutional predistribution of the capital stock itself. Milder sentences in the same grammar include sovereign funds paying citizen dividends, baby bonds endowing each child with a stake at majority, and employee ownership schemes that widen the register of shareholders. The article names these as grammatical forms, not as endorsements, and their evaluation belongs to policy analysis rather than to the theory of exploitation. What belongs here is the observation that the grammar exists, that it is spoken fluently by people who have never heard of the withdrawal criterion, and that its fluency is the relocation’s most diffuse legacy.

The grammar also clarifies an old argument within socialist strategy. The classical tradition’s suspicion of distributive reform, its insistence that rearranging income leaves the relations of production intact, was grounded in the classical location of the wrong. If the wrong is the extraction inside production, then transfers are indeed cosmetic, since the extraction continues beneath them. Roemer’s relocation reverses the valence of one kind of reform while preserving it for another. Transfers of income remain superficial on his account too, since they leave endowments untouched. But reforms of endowment, of the ownership structure itself, are on his account the opposite of cosmetic: they strike the phenomenon at its foundation, more directly than any reorganisation of the labour process could. The strategic map is thereby redrawn. Where the classical map ranked reforms by their proximity to production, the relocated map ranks them by their proximity to the asset distribution, and measures that once looked like the mildest tinkering, inheritance reform above all, acquire foundational significance, while measures that once looked revolutionary, the seizure of workplaces by their occupants, are reclassified as changes in management that leave the exploitation structure’s basis where it stood, unless the seizure changes who owns what. Readers do not have to accept the redrawing to learn from it. Seeing the same reform ranked oppositely by two maps is the fastest way to understand what the maps are maps of.

Why does the relocated account rank inheritance reform so highly?

Because inheritance is the mechanism by which asset inequality reproduces itself across generations. A theory that locates exploitation in the starting distribution must treat the rules governing the transmission of starting positions as foundational, so the taxation and limitation of inherited wealth moves from the periphery of reform to its centre.

The policy grammar has its own critics, and one criticism completes the article’s symmetry. Predistributive reforms, however foundational in the relocated theory, are politically the hardest reforms to enact and to sustain, since they confront concentrated ownership directly rather than skimming its income. A strategy that ranks reforms by theoretical depth may therefore rank them inversely by feasibility, and the classical tradition’s preference for organising at the point of production was, among other things, a judgement about where the exploited could be gathered and moved to act. The relocated theory has no comparable account of agency: its coalitions are analytical constructions, and the question of who would enact the equalisation of endowments, against the resistance of those the equalisation dispossesses, is one the framework poses in its models and cannot answer with them. Roemer’s later work on political economy, on democracy and on the competition between political programmes, addresses the agency question in a different idiom, and the address is partial. The gap is recorded here as the last of the account’s silences. It is the silence about power in the political sense, and it rhymes with the first silence, about power in the workplace. A theory that began by bracketing domination inside production ends by bracketing it inside politics as well, and the bracketing is consistent, principled, and costly in the same way at both ends.

Fidelity and reconstruction: whose Marx is it?

Every reconstruction provokes a question about fidelity, and Roemer’s provoked it in an acute form because the reconstruction’s conclusions, the dispensability of the value theory above all, struck many readers as conclusions Marx would have repudiated. The fidelity question divides into two, and the division dissolves much of the heat. The first question is exegetical: do the classical texts support the relational reading, the distributional reading, or both? The honest answer, as the earlier discussion of the tradition’s purposes suggested, is both, in different passages and with different emphases. The texts contain the analysis of the working day, saturated with the phenomenology of direction and endurance, and they contain the distributive passages, the critique of equal right in the Gotha programme critique, the insistence that distribution relations are the reverse side of production relations. A reconstruction that develops one strand is not overturned by the other strand’s presence. The second question is methodological: what does a living tradition owe its founding texts? Roemer’s practice answers that it owes them its problems and its standards of argument, not its conclusions, and that a conclusion the founder would have repudiated may still be the tradition’s conclusion if the arguments compel it. His critics’ practice answers that a tradition is constituted by its conclusions as much as by its problems, and that a reconstruction which reverses the founder’s central theoretical commitments has founded a new tradition under an old name. The dispute is not resolvable by scholarship, since it is a dispute about what traditions are. It is resolvable, for practical purposes, by labelling, and the literature’s eventual settlement, analytical Marxism as a named school with Roemer as a named founder, is exactly a labelling settlement: the reconstruction acquired its own name, and the old name’s custody ceased to be the prize.

Is Roemer’s account faithful to the classical texts?

In the exegetical sense, partially, as every reconstruction is. The texts sustain both a relational and a distributional reading, and Roemer develops the second while bracketing the first. In the methodological sense, the question divides his readers, since his practice treats the tradition as owing its founder problems and standards rather than conclusions.

One consequence of the labelling settlement deserves emphasis for students navigating the literature. Because the school named itself, its works can be read as a self-contained corpus with its own internal debates, the Cohen and Roemer exchanges, the Wright generalisations, the Elster methodological audits, without constant reference back to the question of Marx’s intentions. The corpus rewards that reading. Its debates are genuine, its results are checkable, and its prose, at its best, has a clarity the founding texts rarely attempted. The cost of the settlement is visible from outside the corpus: the questions the school bracketed, the phenomenology of work, the theory of social form, the critique of the commodity, continued to be worked on by other hands in other idioms, and the two literatures, the analytical and the form-analytic, have largely ceased to read each other. A student of the whole field must therefore read in two traditions to recover one subject, and articles in this series, which are built to the classical architecture while reporting the analytical reconstructions in full, are one attempt to keep the two literatures in a single room. The present article’s verdict, divided between the distributional success and the relational failure, is what the view from that single room looks like when it is honest.

What becomes of surplus value?

A relocation this thorough rearranges the whole conceptual furniture of the tradition, and the fate of surplus value under it is worth a section, since students often assume the concept simply vanishes. It does not vanish. It is demoted. In the classical architecture, surplus value is the load-bearing category: the magnitude whose production explains profit, whose extraction defines exploitation, and whose appropriation constitutes the capitalist class. In Roemer’s architecture, the surplus product, the physical excess of output over the inputs consumed in producing it, remains a fact of any productive economy, and its distribution remains a central object of analysis. What disappears is the category’s moral and explanatory privilege. The surplus is a pie whose division is governed by the asset distribution and the bargaining it structures, and the division’s injustice, where it is unjust, is inherited from the distribution’s injustice rather than generated by a special mechanism of extraction. Surplus labour, correspondingly, becomes one accounting of the surplus among several, the steel accounting being another, and the choice among accountings is a matter of analytical convenience rather than of moral insight. The demotion is the relocation viewed from the side of the tradition’s vocabulary: the words survive, the architecture that gave them rank does not.

The demotion has a clarifying consequence for a debate that long predates Roemer, the debate about whether exploitation requires that the exploited produce the surplus. In some classical formulations, the exploited class is specifically the productive class, and unproductive workers, however badly paid, stand outside the exploitation relation strictly defined. The withdrawal criterion dissolves the distinction’s basis: a coalition’s status depends on its endowments and its counterfactual, not on whether its members’ labour is classified productive, and a badly paid service worker with no assets is exploited in exactly the same sense as a factory hand with no assets. Many readers experience this dissolution as a gain in sanity, since the productive and unproductive distinction had generated scholastic disputes disproportionate to its insight. Others note that the distinction encoded a claim about where the surplus originates, and that dissolving it abandons the claim without refuting it. Roemer’s position, once more, is that the claim about origins was the accounting correspondence in disguise, and that the generalised theorem has already dealt with it. The exchange is a miniature of the whole controversy, and it is left here, deliberately, as the reader’s exercise: the materials for both verdicts are in the preceding sections, and the defeater discipline applies to whichever verdict the reader reaches.

Does the withdrawal account distinguish productive from unproductive labour?

No, and the omission is principled. Status under the criterion depends on endowments and the withdrawal comparison, not on the classification of the labour a coalition performs, so the badly paid service worker and the factory hand with identical asset positions receive identical verdicts, which the account counts as a clarification.

There is a companion demotion to record, that of the rate of exploitation. The classical ratio of surplus to necessary labour gave the tradition a number, comparable across industries and epochs, around which empirical and historical work could organise. The withdrawal framework offers no successor number of comparable portability. A coalition’s gain from withdrawal can be computed inside a model, but the computation is model-specific, and the literature has not converged on a standard measure that travels between studies the way the classical ratio travelled between texts. The absence matters for the account’s scientific standing in a way its defenders concede too rarely. A theory that dissolves a measurable magnitude and replaces it with a family of model-relative computations has traded empirical convenience for conceptual clarity, and the trade should be entered in the ledger beside the others this article has kept: generality bought with abstraction, tractability bought with silence about production, and now measurability surrendered in the passage. The ledger’s balance is the verdict already delivered. The relocation is a great clarification and a real impoverishment, and the two descriptions are of the same act.

Cooperatives, the self-employed, and the middle of the distribution

The correspondence principle’s middle case, the agent who works her own assets and neither hires nor sells labour, is theoretically quiet and empirically enormous, since the self-employed, the peasantry of the classical texts, the artisan, the small shopkeeper, and the independent professional all live there. The withdrawal criterion pronounces such agents neither exploiters nor exploited in the model’s central case, and the pronouncement organises a set of questions that the classical account handled through the category of the petty bourgeoisie and its vacillations. Roemer’s framework translates the vacillation into arithmetic. The independent producer’s status is poised: let her accumulate assets and begin to hire, and she crosses into the exploiting class; let her assets fail and force her into the labour market, and she crosses into the exploited. The crossings are continuous in the asset dimension, and the framework predicts, what the classical narrative also describes, a stratum in permanent transit, replenished and drained by the fortunes of accumulation. The translation loses the narrative’s political psychology, the stratum’s fears and resentments, and gains a precision about the stratum’s boundaries: they are where the returns to hiring and to selling cross the returns to independence, at given prices, for given endowments.

Worker cooperatives pose the sharper test, since they are the classical tradition’s own institutional remedy, the associated producers taking possession of their workplaces. Under the withdrawal criterion, a cooperative’s members, owning their firm’s assets jointly, are independent producers in the model’s sense when the cooperative is self-contained, and the criterion finds no exploitation among them. But the cooperative exists inside an economy, and the criterion’s economy-wide application asks the further question: what is the cooperative members’ coalition position relative to the whole asset distribution? A prosperous cooperative in an economy of concentrated ownership may constitute, collectively, an asset-holding group that would lose from an equalising departure, and its members then occupy, by the criterion’s lights, a position analogous to exploiters relative to the asset-poor outside, however democratic their internal arrangements. The verdict is uncomfortable and illuminating in equal measure. It captures something real about the politics of cooperative movements, their tendency to become insider institutions defending a stake, which the classical vocabulary of workplace democracy cannot express. And it does so by exactly the move the domination objection protests: the internal relations of the cooperative, the axis on which its members’ lives are actually governed, are set aside, and only the external asset position is read. The case is therefore a perfect examination question, since a complete answer must run both accounts, report both verdicts, and explain why they diverge, and the article leaves it posed in that form for the teachers among its readers.

How does the criterion classify a prosperous cooperative’s members?

By their coalition’s position in the whole economy’s asset distribution, not by their internal arrangements. Jointly owning valuable assets, the members may sit in a coalition that an equalising withdrawal would harm, placing them on the exploiting side of the economy-wide test while their internal democracy leaves the relational account nothing to criticise.

The middle of the distribution also houses the figure the framework handles most delicately, the professional with scarce credentials. Her income is high, her assets may be modest, and her advantage flows from an inalienable endowment, certified skill, rather than from alienable holdings. Under the capitalist test she may be exploited or neutral depending on her asset coalition. Under the socialist test, which imagines inalienable endowments equalised, she is among the exploited’s beneficiaries, a recipient of the skill premium the test indicts. Wright’s class analysis, which treats credentials as an exploitation-relevant asset in their own right, can be read as an attempt to bring this figure inside a single framework without the two-step of successive tests. The attempt’s difficulties, the multiplication of dimensions and the dilution of the correspondence, were noted earlier. The figure herself remains, in every framework, the citizen whose position the twentieth century’s class theories were least built to see, and the Roemerian taxonomy at least has a named cell waiting for her, which is more than the classical two-class schema could say.

The state of the question

The debate this article has chronicled has reached a stable configuration in the literature, and describing the configuration is a better close to the exposition than any pretence that the debate continues at its former temperature. The withdrawal criterion and the correspondence principle are canonical: they appear in the reference works, the graduate syllabuses, and the textbooks of Marxian economics and of political philosophy, stated in the form this article has used. The generalised theorem is unreversed: no reply has restored labour’s privilege on formal grounds, and the replies that matter have conceded the formal ground and contested its significance. The domination account has been built out into a literature of its own, with the relational axis now possessing the formal and institutional development it once lacked. The two-axis map is, in effect, the field’s working settlement, honoured even by partisans who describe it as a staging post. Roemer’s own trajectory, from the exploitation theory through market socialism to equality of opportunity, is studied as a single arc, and the arc’s unity, the recurring relocation of moral attention from conduct to background, from transactions to the distributions that frame them, is recognised as one of the most consistent intellectual projects in the period’s political economy.

What remains genuinely open can be stated in three questions, and the article ends its exposition by bequeathing them. First, the construction invited by this article’s defeater: can a withdrawal-style criterion be formulated that incorporates a production relation without surrendering the isomorphism and generality results, and what exactly would the incorporation cost? The question is precise enough to work on, which is the defeater discipline’s purpose. Second, the measurement question: can the distributional axis be given an operational measure of exploitation that travels between studies as the classical ratio once did, or is model-relativity the permanent condition of the relocated concept? Third, the agency question: what political sociology corresponds to the relocated theory, who enacts endowment equalisation and under what banners, in a framework whose coalitions are analytical rather than organised? None of the three questions is rhetorical. Each marks a place where the construction’s successors will either build or acknowledge a boundary, and a reader who carries the questions away has the article’s argument in its most useful form, as a set of tools with their tolerances stamped on them.

Which results of the construction stand unreversed?

The correspondence between class position and exploitation status under the model’s conditions, and the generalised theorem’s dissolution of labour’s formal privilege. Replies in the literature concede both results and contest their significance, relocating the debate to what the results mean rather than whether they hold, which is the configuration this article’s verdict reflects.

A last word on method, addressed to the reader who has followed the whole argument. The relocation controversy is the finest available training ground for a skill that every other article in this series presupposes: the skill of watching a concept being rebuilt and of keeping the accounts while it happens. What was the concept’s work? What does the rebuilt version still do? What has it declined, openly or otherwise? What would change the verdict? A reader who can run those four questions over Roemer’s exploitation can run them over any reconstruction in any field, and will find that the questions, unlike the answers, never go out of date.

Among the egalitarians: the criterion’s philosophical cousins

Roemer’s criterion did not develop in a philosophical vacuum, and placing it among its egalitarian contemporaries sharpens its profile. The early nineteen eighties were the years in which liberal egalitarianism rebuilt itself around the distribution of resources and opportunities: Rawls’s theory was a decade old and under reconstruction by its critics, and Ronald Dworkin’s equality of resources, proposed in 1981, imagined an auction in which equal bundles of resources are bid for with equal purchasing power, followed by insurance markets against bad luck in talents. The family resemblance to Roemer’s constructions is not accidental and not imitation. Both projects ask what an equal starting position would look like and measure actual arrangements by their distance from it. Both treat talents as a problem for equality rather than as a natural entitlement. Both were built by theorists trained to state their constructions precisely enough to be refuted. The differences are equally instructive. Dworkin’s auction is a device for defining equality itself, a pattern a distribution should exhibit. Roemer’s withdrawal is a device for classifying positions within a distribution acknowledged to be unequal, a diagnostic rather than a pattern. And Roemer’s device is yoked to a class analysis and to a tradition’s conclusions about capitalism, where Dworkin’s is yoked to a liberal theory of rights. The cousins share a grandfather, the idea that justice concerns the fairness of the starting gate, and their differences are the differences between the traditions that raised them.

The comparison also locates Roemer’s distinctiveness within egalitarianism itself. Most egalitarian theories of the period were theories of the pattern justice requires, and their practical content was exhausted by stating the pattern. Roemer’s theory, because it grew from Marxian soil, carried an explanatory ambition the others lacked: it wanted not only to say which distributions are unjust but to explain the social structure, classes, firms, markets, that unjust distributions generate, and to derive the pattern of complaint from the structure rather than merely deploring it. That ambition is the correspondence principle’s raison d’être, and it is why the exploitation theory, alone among the period’s egalitarian constructions, comes with a sociology attached. Egalitarians who found the sociology compelling had to take the exploitation framework seriously. Marxists who found the egalitarianism compelling had to take distributive ethics seriously. The cross-traffic the construction generated, in both directions, is a large part of its historical importance, larger perhaps than any of its individual theorems.

How does the withdrawal test differ from Dworkin’s auction?

In purpose. The auction defines the pattern an equal distribution would exhibit, a standard for arrangements. The withdrawal test classifies positions inside an existing unequal distribution, a diagnostic that sorts coalitions into exploited and exploiting by their distance from the equal division the auction imagines being enacted.

The traffic had a terminus that neither tradition predicted in 1982. By the century’s end, the most influential egalitarian debates concerned opportunity, responsibility, and luck, and Roemer was a central figure in them, while the most influential Marxian debates concerned recognition, culture, and the state, territories his framework never entered. The exploitation theory sits at the fork, the last major work in which the two traditions were one conversation. Reading it with the fork in view gives the article’s subject its final dimension. The relocation was not only an argument about a concept. It was the point at which a tradition’s best economist walked its central normative claim across a border into a neighbouring discipline, and the neighbouring discipline, to its credit, made him a citizen rather than a tourist. What the home tradition lost in the crossing, and what it gained, is the question this series exists to keep open, and it is fitting that the question should be unanswerable in any terms simpler than the two-axis map the controversy produced.

Reading Roemer: the texts in their order

A short guide to the primary texts belongs in an article of this kind, since the secondary literature’s compressions are the source of half the misreadings corrected above. Begin with Free to Lose, the 1988 introduction, where the theory is stated for the general reader with its ethical conclusions in the foreground and its formalism in the background. A reader who masters that book possesses the criterion, the correspondence in outline, and the farewell to the value theory, in the author’s own most accessible prose. Proceed then to the essays on exploitation and justice gathered in Egalitarian Perspectives in 1994, where the philosophical joints of the construction, the benchmark question, the relation to distributive ethics, the replies to critics, are worked at in detail. Turn then to the 1982 foundation itself, reading it for the results as stated and the architecture of the taxonomy, and dipping into the proofs as equipment allows. Read A Future for Socialism, also 1994, last among the major works, since the coupon design presupposes the diagnosis and reads, after the theory, as its institutional corollary rather than as a free-standing proposal. The equality of opportunity writings belong to a later chapter of the author’s career and to other articles’ territory, but the reader who reaches them will recognise the root system at work under new growth.

Two habits make the reading profitable. The first is to track the definitions as they shift between texts, since Roemer revised his formulations as objections arrived, and a criticism aimed at the 1982 statement may be answered by the 1988 restatement or conceded in the essays. The second is to read the critics in the order the article has presented them, Cohen for the unfreedom and the producer’s claim, Wright for the pluralisation of assets, Vrousalis for the domination reconstruction, since the critics are not a chorus but a sequence, each answering the construction at a different joint. A seminar that pairs each Roemer text with its appointed critic will find that the debate teaches itself, and that the students’ verdicts, when they come, divide along the two axes with a fidelity that confirms the map this article has drawn.

The test across borders: coalitions in a world economy

Nothing in the criterion confines its coalitions within national frontiers, and applying the test across borders produces some of its most provocative and least settled readings. Define the relevant economy as the world, parcel its alienable productive wealth by heads across the whole human population, and test the coalition of a poor country’s workers against departure with their global per capita portion. The verdict, in any plausible model, is exploitation on a scale that dwarfs the domestic readings, since the global maldistribution of capital dwarfs the domestic one. The exercise clarifies, by enlarging it, the criterion’s essential character. The verdict does not depend on any transaction between the poor country’s workers and the rich world’s citizens, on any trade relation, any investment flow, any history of extraction between the specific parties. It depends only on the global endowment map and the counterfactual parcelling. A classical theorist finds the reading unmoored, since exploitation without an exploiting relation, without surplus flowing between the parties named, seems to accuse the rich world’s citizens of a wrong they may never have transacted with the coalition in question. A Roemerian finds the reading the framework’s honesty at planetary scale: if the wrong is the endowment map, then the map’s global extent is the wrong’s extent, and the absence of a transaction is the absence of a mechanism the theory never required.

The border case exposes, in addition, how much the criterion depends on the definition of the economy, a dependence domestic applications conceal. Which assets enter the parcelling when the economy is the world? Productive wealth is held under national legal systems, denominated in national currencies, protected by national courts, and a global per capita share is a claim on a stock that no global institution exists to divide. The counterfactual’s coherence, already strained by indivisible installations inside one country, strains further across jurisdictions, and the strain is informative rather than merely embarrassing. It shows that the criterion presupposes a political unit within which endowments could in principle be redivided, a community whose members acknowledge the parcelling as a meaningful standard. Where that presupposition holds, inside a state with a tax system and a law of inheritance, the criterion has institutional traction: the counterfactual it imagines is one the state’s own machinery could approximate. Where the presupposition fails, between states, the criterion continues to return verdicts, but the verdicts have changed their character from measurements against an actionable standard to judgements against a moral one. Roemer’s own work stayed largely within the presupposition’s home territory. The framework’s export across borders, by development economists and by theorists of global justice, is a chapter written by other hands, and the chapter’s unsettled state is the honest condition of a tool being used beyond its design tolerances.

Why do cross-border applications change the verdict’s character?

Because the parcelling counterfactual presupposes a political community able in principle to redivide endowments. Inside a state, that community exists and the verdict measures distance from an actionable standard. Between states it does not, so the same computation expresses a moral judgement about the global endowment map rather than a measurement any institution could act on directly.

The border discussion returns the exposition, by a long route, to the article’s opening observation. The relocation began as a move inside Marxian economics, from the factory floor to the asset register. Followed consistently, it ends as a claim about the moral geography of endowments wherever endowments are found, indifferent to the frontiers and the firm boundaries that organised the classical concept’s world. That indifference is the relocation’s grandeur and its exposure in a single property. A concept that can pronounce on any distribution can be accused of pronouncing on none in particular, of having traded the classical concept’s grip on a specific mechanism for a universal solvent in which every inequality dissolves into the same verdict. The accusation is the emptied-concept objection at its largest scale, and it meets the same answer at that scale: the solvent is not universal, since it leaves domination, need unconnected with endowments, and the historical manner of acquisition untouched and visible beside its readings. A map is not falsified by features it never claimed to chart. It is refuted only by mischarting the features it claims, and the long argument of this article has been an audit of the relocation’s charting of its chosen feature, the endowment distribution and the class structure it generates. The audit’s finding stands as the verdict delivered it: the charting is exact, the territory charted is real, and the territory is not the whole country.

Why the account matters to readers who reject it

An exposition that has divided its verdict owes the reader a final account of why the losing half of the argument, whichever half the reader takes that to be, was worth the hearing. Consider first the reader who leaves persuaded by the domination objection, holding that a concept of exploitation without directed labour in it has been emptied past recognition. That reader still owes Roemer three debts. The first is the theorem: after the generalised result, no defence of the classical account can rest on the accounting correspondence, and every serious defence written since has been written in the space the theorem cleared, on grounds of relation and human character that the theorem forced into the open. The second debt is the correspondence: any relational theory must now explain why the relations it prizes are populated with such regularity by the endowment distribution, and a domination account that cannot say why the asset-poor are predictably the dominated has a hole where its sociology should be. The third debt is methodological: the relocation set a standard of explicitness, premises on the table and defeaters attached, against which the relational literature has sharpened itself, Vrousalis’s construction being unthinkable in its achieved rigour without the target Roemer provided.

Consider then the reader whom the relocation persuades, holding that distribution was the phenomenon all along. That reader owes the critics a parallel set of debts. The domination literature has established, beyond serious reply, that something the tradition cared about escapes the distributional net, and a Roemerian who proceeds as if nothing escaped will misdescribe workplaces, households, and every institution where command does the work that ownership does elsewhere. Cohen’s unfreedom argument stands as a permanent reminder that the language of choice, which the models cannot avoid, systematically flatters the arrangements it describes, and that the flattery must be corrected by hand wherever the models are applied. And the feminist and labour process literatures have shown that endowments themselves are made inside relations, so that the distribution the criterion takes as its starting point is always also a result, a sediment of prior direction and appropriation, which a complete social theory must excavate rather than merely measure. The two readerships, in other words, leave the controversy needing each other, and the controversy’s product is not a winner but a division of labour with the border surveyed. Surveys of that quality are rare in the human sciences, and they are what the analytical episode, at its best, contributed to the tradition it reconstructed and to the disciplines it argued with.

The relocation restated

The argument of this article can be compressed, at the end of its exposition, into a sequence of claims each of which has been established or contested in the sections behind it. Exploitation, in the classical account, is a process inside production, the extraction of surplus labour, and its analysis requires the categories of value. Roemer redefined exploitation as a property of endowment patterns: a coalition is exploited when departure carrying the economy’s alienable assets, divided equally by heads, would leave it better off. Under stated conditions, the classes generated by an asset distribution coincide exactly with the exploitation classes the redefinition sorts out, so the tradition’s pairing of class and exploitation survives as a theorem. The formal argument that was thought to privilege labour in this story privileges any basic commodity equally, steel included, so labour’s privilege, if it exists, rests on labour’s human character and not on value accounting. The ethical critique of capitalism is thereby relocated onto the justice of asset distributions, and the labour theory of value becomes dispensable to it. The relocation purchases generality across social forms, tractability of verdicts, and an indictment independent of contested price theory. It pays with the concept’s grip on domination, on the directed and appropriated character of labour inside production, which the redefined concept cannot represent. That judgement carries its condition of defeat: it falls if a successor to the withdrawal test ever separates exploitation from plain inequality of endowments by pointing to a relation that holds between people while production goes on, and no such successor exists in the framework examined here.

Compressed sequences of that kind are dangerous in proportion to their usefulness, and the reader is warned against carrying the compression away without the qualifications the exposition attached to each claim. The correspondence holds under conditions. The theorem addresses a style of argument. The criterion presupposes a benchmark it does not defend. The domination it loses is lost by a framework, not by a word, and other frameworks in the same literature keep it. With the qualifications restored, the sequence is the article’s finding, and it is offered in the series’ manner: as a judgement that knows what would reverse it, on a subject where the pretence of finality has been the only position conclusively refuted. What Roemer built stands as the most rigorous answer the tradition has produced to the question of what exploitation is, if the question is asked about distributions. What his critics preserved stands as the most rigorous reminder that the question can also be asked about relations, and that the tradition asked it that way first. The reader who holds both, and can say which question any given argument is answering, has mastered the material in the only sense mastery has here.

For students and exam candidates

Build the answer around the relocation, and state it in the first paragraph: Roemer derives exploitation from the distribution of productive assets by a withdrawal test, requiring no premise about the labour process. Then give the three results in order, the criterion, the class-exploitation correspondence, and the generalised commodity exploitation theorem, since examiners reward the sequence and penalise answers that reach the theorem without the criterion beneath it. Keep the classical account beside the new one at every step, using the comparison this article’s table fixes, and take the classical account’s own derivation from the series article on the classical concept rather than improvising it. Close with the split verdict and the condition attached to it: mastery of the distributional basis, silence about command inside production, and the verdict on that silence lapses the day somebody rebuilds the test around a relation within production itself. A VaultBook card on the two accounts, kept with the Marxism study notes, is the revision instrument this article recommends for holding the contrast steady. The commonest examination failure on this material is to narrate Roemer as a critic of Marx from outside. He is a reconstructor from inside, and answers that register the intention before assessing the result start a full band ahead.

For teachers and seminar leaders

The withdrawal criterion is the most teachable construction in the analytical literature, because it can be made concrete in a single seminar exercise without any formalism. Divide the room into coalitions, assign endowments, and let each coalition compute whether it would gain from departure with a per capita share; the verdicts, and the arguments about coalition boundaries that erupt while computing them, teach the theory faster than exposition does. Follow the exercise with the steel substitution, which students should be required to construct for themselves before being shown Roemer’s version, since the construction’s simplicity is its persuasion and a student who has built it owns the theorem. Reserve the final session for the strongest objection, the supervised worker and the small asset-holder, and require the class to state Roemer’s answer and its concession before offering verdicts of their own. The defeater exercise transfers well: ask each student to write the sentence that would reverse their verdict, and treat any verdict submitted without one as unfinished. The texts divide naturally across four meetings, Free to Lose for the statement, the 1982 book for the architecture, Cohen and Vrousalis for the two flanks of criticism.

For researchers, historians, and journalists

The research frontier this article leaves open has three named gates: a withdrawal formulation incorporating a production relation, an operational measure of distributional exploitation that travels between studies, and a political sociology of endowment equalisation. Work at any of the three gates should cite the construction precisely, distinguishing the general coalition form of the definition from its class form, since a large share of the secondary literature’s confusion lives in that distinction. Historians will find the criterion’s feudal application and the transition models the most usable parts of the corpus, and should note that the framework’s historical claims are modal, about what was necessary, rather than narrative, about what occurred. Journalists covering ownership policy, citizen funds, inheritance reform, or employee ownership are working inside the relocation’s policy grammar whether they know it or not, and the framework offers them a discipline: name the coalition, specify the asset base, and state what the proposal does to the starting distribution rather than to incomes after the fact. Researchers in the domination literature will find the two-axis map the honest meeting point, and should resist citations that crown either axis the whole of exploitation.

Frequently Asked Questions

Q: How does Roemer define exploitation?

Roemer defines exploitation through a test of hypothetical departure. Take any group: if its members would fare better by leaving the economy while carrying an equal per-person portion of the society’s alienable productive wealth, the group counts as exploited; if departure on those terms would leave them worse off, they count as exploiting. Nothing about hours, effort, or supervision enters the calculation. The phenomenon is fixed entirely by how productive assets are spread across the population before production starts.

Q: What is the withdrawal criterion?

The withdrawal criterion is the hypothetical-departure test that does the defining in Roemer’s system. The analyst picks a coalition, imagines it seceding with a stated endowment, and asks whether its members’ lives would go better outside than inside. For the capitalist variant the seceding group takes an identical slice per head of every alienable asset. Other variants fit other social orders: a feudal group leaves with the holdings it already has, quit of its dues, while a further variant imagines talents themselves shared out evenly. The test diagnoses; it proposes no actual secession.

Q: What is the class exploitation correspondence?

This is the name for Roemer’s proof that two independently built classifications land on the same people. In his model economy, households endowed unequally choose how to employ their time and property. Those rich in assets find hiring others most rewarding; those with almost none find selling their own effort most rewarding; a middle band simply works its own property. The theorem states that the first band are exploiters and the second exploited under the departure test, so market behaviour and exploitation status trace a single division.

Q: What is the generalised commodity exploitation theorem?

Roemer’s generalisation of the fundamental Marxian theorem runs the older proof with a different input in the starring role. Reckon every good’s value by the steel used up in making it, counting steel used to make the steel, and a surplus of steel over replacement emerges exactly when profits are positive, paralleling the surplus of labour over the wage bundle. Since steel can suffer no injustice, the parallel shows that correspondences of this kind cannot by themselves mark labour out as the wronged factor.

Q: Does the argument single out labour?

It does not, and the failure is the theorem’s lesson. The reasoning that moves from positive profit to an exploited factor works for steel, energy, or any basic input, because the proof leans only on features all such inputs share: being consumed in production, being embodied indirectly in other goods, needing replacement. If labour is to be privileged, the privileging premise must concern what labourers are, namely persons, and moral premises about persons do the work directly, without value accounting as an intermediary.

Q: What do Marxists object to in this account?

Critics inside the Marxist tradition, Cohen and Vrousalis among them, press three charges. The first says the relocation trades a relation for a distribution, discarding the directed character of work that the concept existed to name. The second says the model’s choosing agents disguise compulsion, since owning nothing but one’s capacity to work is itself a coercive condition, as Cohen’s work on collective unfreedom argues. The third points to verdicts the criterion’s author himself calls counterintuitive, where comfort and exploitation coincide or hardship escapes the label.

Q: Does the account capture domination?

It does not, on the verdict this article defends. Pacing, monitoring, subjection to a foreman’s authority, and the surrender of the product to whoever directs the work appear nowhere among the framework’s premises, so command over a worker registers only through the asset positions behind it. Two people with matching endowments receive matching verdicts however different their experience of command. Roemer’s defenders answer that command is a separate wrong deserving its own concept, and that keeping the two apart is clarification rather than loss.

Q: Why does Roemer abandon the labour theory of value?

His reason is that the ethical case survives the theory’s removal, so the removal costs nothing he values. One of his theorems shows that embodied-labour accounting cannot privilege labour over other inputs; his criterion then rebuilds exploitation verdicts out of asset holdings alone, never mentioning values. A critique of capitalism that rests on contested price theory can be blocked at that foundation, while a critique resting on unjust holdings cannot. Roemer stated the conclusion openly: the value theory should be retired as the critique’s foundation.

Q: What is coupon socialism?

Coupon socialism is the ownership constitution Roemer drafted in A Future for Socialism to abolish exploitation in his sense while keeping markets. Each adult holds an equal book of coupons, spendable solely on shares in firms, through competing mutual funds. The coupons may not be sold for cash, bought up from others, or passed on at death; estates surrender them for redistribution to the incoming generation. Dividends are ordinary spendable income. Firms keep professional managers and market discipline, but the capital stock’s ownership is equalised by constitutional design rather than by a founding confiscation.

Q: Is the withdrawal criterion counterintuitive?

By its author’s own admission, in a run of cases, yes. Someone who lived idly while equals saved, and who must later sell her labour to them, is not exploited on the test, though her job matches any other. Groupings drawn differently around the same people flip verdicts. And when talents are hypothetically shared out, the highly skilled come out exploited by the rest. Roemer printed these cases himself and absorbed them, arguing that a test faithful to distributional grievances will sometimes part company with the intuitions attached to the word.

Q: How does the relocation grade the new account of exploitation?

The grade this article returns is split along the line the comparison table draws. Everything the account promises on the distributional side is delivered: classes fall out of endowments, verdicts compute, and labour’s formal privilege dissolves. Everything the classical concept carried about life inside production, direction, endurance, appropriation, falls outside the new concept’s reach. So the relocation is recorded as a success about exploitation’s basis and a failure about exploitation’s full extent, with the failure reversible only by a reformulation of the test that builds a production relation into it.

Q: What should a student separate between exploitation in production and inequality of assets?

An answer should treat the two as rival answers to the question where the wrong sits, and never let them blur. One candidate location is a process: surplus labour extracted under another’s direction during production. The other is a starting condition: holdings of productive wealth spread unequally before anyone clocks in. Roemer’s claims all belong to the second location, the classical claims to the first, and most confusion in student work comes from borrowing a premise from one side to support a conclusion on the other. Naming the location in play is the discipline examiners reward.

Q: Why does the withdrawal test not need the working day?

The comparison the test performs is between two positions of a whole coalition, its standing inside the economy and its standing after a hypothetical departure with equalised holdings. Neither term of that comparison describes a shift at a bench or a spell at a machine. How long anyone works, how hard, and under whose eye are facts the classical account needs and this one never consults, which is precisely what allows a single instrument to judge feudal dues, wage labour, and skill premiums in one vocabulary.

Q: How does the two accounts compared table order the definitions?

Its five rows put the same questions to both definitions in the same order, so the reader compares like with like instead of weighing one account’s best feature against the other’s worst. The opening rows ask where the phenomenon lives and what must hold before either account will pronounce; the middle rows ask what each explains and what remedy each implies; the closing row asks what slips through each account’s net. That last row does the verdict’s work, since the relocation’s success and failure are both recorded there.

Q: What separates being exploited from being badly off?

Hardship and exploitation part company at the question of remedy. A condition counts as exploitation, in the relocated sense, when sharing the economy’s productive wealth equally would lift it; hardship from illness, accident, or bereavement under an already equal parcelling fails that test and belongs to other categories of injustice or misfortune. The parting also runs in reverse, since a prosperous employee whose asset-poor coalition would gain from the hypothetical departure is classified as exploited despite her comfort. Condition alone never settles the classification; the counterfactual does.